Book Vendor Records in the OCLC Database 10 Boon Or Bane? Laura D

Library Resources & Technical Services

ISSN 0024-2527 January 2001 Volume 45, No. 1

ARTICLES

Using Citation Analysis to Pursue a Core Collection of Journals for Communication Disorders 3 Steve Black Book Vendor Records in the OCLC Database 10 Boon or Bane? Laura D. Shedenhelm and Bartley A. Burk The Effect of Interface Design on Item Selection in an Online Catalog 20 David H. Thomas The Year’s Work in Cataloging, 1999 47 Amy K. Weiss and Timothy V. Carstens

FEATURES

Book Reviews 59 Margaret Rohdy, Editor From Carnegie to Internet2, reviewed by Bob Persing American Library From Gutenberg to the Global Information Infrastructure, Association reviewed by Sheila S. Intner Association for Library The Future of Cataloging, reviewed by Sheila S. Intner Collections & Technical Progress in Visual Information Access and Retrieval, Services reviewed by Sara Shatford Layne Visit LRTS online at Social Life of information, reviewed by Michael Gorman www.ala.org/alcts/lrts. For current news and Instructions for Authors 65 reports on ALCTS activities, see the ALCTS Newsletter Online at Instructions for Book Reviewers 67 www. ala.org/ alcts/alcts_news. Index to Advertisers 68 EDITORIAL BOARD

Editor and Chair of the Editorial Board John Budd University of Missouri

Assistant Editors Ex-Officio Members Elizabeth E. Cramer, Appalachian State University Elaine Yontz, University of Southern Florida Prudence W. Dalrymple, Dominican University Delegate of the Chair, Council of Regional Groups Jeanne M. Drewes, Michigan State University Karen Muller Executive Director, ALCTS Martin M. Kurth, Cornell University Miriam Palm, Stanford University, emerita Margaret A. Rohdy, University of Pennsylvania Editor, ALCTS Newsletter Online Jane Treadwell, Emory University Barbara A. Winters, University of Georgia

Library Resources & Technical Services (ISSN 0024-2527) is published quarterly by the American Library Association, 50 E. Huron St., Chicago, IL 60611. It is the official publication of the Association for Library Collections & Technical Services, a division of the American Library Association. Subscription Price: to members of the Association for Library Collections & Technical Services, $27.50 per year, included in the membership dues; to nonmembers, $55 per year in U.S., Canada, and Mexico, and $65 per year in other foreign countries. Single copies, $15. Periodical postage paid at Chicago, IL, and at additional mailing offices. POSTMASTER: Send address changes to Library Resources & Technical Services, 50 E. Huron St., Chicago, IL 60611. Business Manager: Karen Muller, Executive Director, Association for Library Collections & Technical Services, a division of the American Library Association. Send manuscripts to the Editorial Office: John Budd, Editor, Library Resources & Technical Services, University of Missouri, 221M Townsend Hall, Columbia, MO 65211; (573) 882-3258; fax: (573) 884-4944; e-mail: [email protected]. Advertising: Bill Coffee, c/o Benson, Coffee & Associates, 1411 Peterson Ave., Park Ridge, IL 60068; (847) 692-4695; fax: (847) 692-3877. ALA Production Services: Troy D. Linker, Kevin Heubusch; Ellie Barta-Moran, Angela Hanshaw, and Karen Sheets. Members: Address changes and inquiries should be sent to Membership Department—Library Resources & Technical Services, 50 E. Huron St., Chicago, IL 60611. Nonmember subscribers: Subscriptions, orders, changes of address, and inquiries should be sent to Library Resources & Technical Services, S & S Computer Services, Inc., 434 W. Downer, Aurora, IL 60506. Library Resources & Technical Services is indexed in Library Literature, Library & Information Science Abstracts, Current Index to Journals in Education, Science Citation Index, and Information Science Abstracts. Contents are listed in CALL (Current American—Library Literature). Its reviews are included in Book Review Digest, Book Review Index, and Review of Reviews. Instructions for authors appear on p. 65–68 of the January 2001 issue and on the LRTS Web page at www.ala.org/alcts/lrts. Copies of books for review should be addressed to Margaret Rohdy, Book Review Editor, Library Resources & Technical Services, Van Pelt-Dietrich Library Center, University of Pennsylvania Libraries, 3420 Walnut St., Philadelphia, PA 19104-6206 (e-mail: [email protected]). © American Library Association 2001 All materials in this journal subject to copyright by the American Library Association may be photocopied for the noncommercial purpose of scientific or educational advancement granted by Sections 107 and 108 of the Copyright Revision Act of 1976. For other reprinting, photocopying, or translating, address requests to the ALA Office of Rights and Permissions, 50 E. Huron St., Chicago, IL 60611. The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences—Permanence of Paper for Printed Library Materials, ANSI Z39.48-1992. ∞ Publication in Library Resources & Technical Services does not imply official endorsement by the Association for Library Collections & Technical Services nor by ALA, and the assumption of editorial responsibility is not to be construed as endorsement of the opinions expressed by the editor or individual contributors. 45(1) LRTS 3

Using Citation Analysis to Pursue a Core Collection of Journals for Communication Disorders Steven Black

A citation analysis from a purposive sample of two leading journals is employed to build a tentative core collection of journals in communication disorders. A core collection is defined for this study as those journals that provide 80% of the sample’s article citations. The bibliometric concept of “success-breeds-success” is reviewed, and its application to this sample of journals is quantified. The special problems of defining a core collection in a multidisciplinary field are discussed. Data is also provided on the types of publications cited, and the age distribution of cited journals.

he field of communication disorders encompasses the study of impairments Tand disabilities in speech, language, and hearing. The American Speech- Language-Hearing Association (ASHA), founded in 1925, accredits practitioners, programs, and college and university graduate programs in communication disorders, and has 97,062 members (American Speech-Language-Hearing Association 2000). ASHA’s Council of Academic Accreditation (2000) has accredited 253 graduate programs in the United States in speech-language pathology and audiology (www.asha.org/students/caa_programs/caaprog.htm). Communication disorders draws from multiple disciplines, yet it is clearly a discipline in its own right. It satisfies Kuhn’s (1996) characteristics of a discipline, having its own specialized journals, an established professional society, and a special place in the academic curriculum. Accredited professionals, students in programs, and the libraries serving them need to access literature from a wide range of fields. The broad, multidisciplinary nature of communication disorders is demonstrated by the works cited in articles published in journals such as the Journal of Speech, Language and Hearing Research and the Journal of Communication Disorders. Works cited in these communication-disorders journals draw from publications in many fields, including audiology, neurology, linguistics, medicine, physiology, psychology, psychiatry, education, and special education. Steve Black ([email protected]) is Identifying the most important journals helps professionals use their time Reference, Instruction, and Serials efficiently, and helps libraries make the most of their budgets. Typically, an aca- Librarian, Neil Hellman Library, the demic library will develop a core collection of journals in each of the disciplines College of Saint Rose, Albany, New York. taught in the institution’s programs, and also subscribe to additional journals that Manuscript received January 27, 2000; strengthen and enhance the core collections. But defining a core collection is not accepted for publication May 15, easy. Many journals are used in more than one discipline, and titles often cease, 4 Black LRTS 45(1)

merge, change names, or are newly created. Also, the qual- hath not shall be taken away even that which he hath.” ity of individual journals changes over time, so reputation Merton (1968) first described the Matthew Effect in terms may overestimate or underestimate the current quality of a of the disproportionate amount of attention given to the journal. work of the most prominent scientists. Models of journal citation distribution models all show a pattern of success- breeds-success, although there is no standard mathemati- Citation Analysis cal formula for the skewed distribution of cited works (Oluic-Vukovic 1997). Bensman and Wilder (1998) Citation analysis is an established method for identifying the described in some detail this and related expressions of the leading journals that belong in a discipline’s core collection. skewed distribution of citations, including “Bradford’s law Citation analysis is the systematic, quantitative study of of scattering,” “Garfield’s law of concentration,” and the works cited. It is part of the broader field of bibliometrics, negative binomial distribution. Other success-breeds-suc- the application of mathematical and statistical methods in cess “laws” have been devised by Pareto, Lotka, and Zipf the study of the use of documents and publication patterns (Oluic-Vukovic, 1997). (Osareh 1996). Osareh (1996) reviewed a body of citation All expressions of the distribution have a common analysis research used to rank publications according to model of concentrated use in a small portion of the whole, their importance, to identify core collections, to measure the and a large, gradually diminishing portion that receives little impact of publications, and to study subject interrelation- use. According to Bradford (1953), the numbers of periodi- ships. Despite potential problems with citation analysis data cals in a nucleus of journals and succeeding zones will follow discussed below, measuring cited use in professional jour- the formula l:n:n2. Unfortunately, one has to perform a cita- nals is a well-established, objective, quantitative method for tion analysis of the journals of a particular discipline to measuring the value of published literature. determine the n in Bradford’s law of scattering. A core col- Citation analysis is based on the premise that citations lection can be determined by following Bradford’s law, but to literature more or less accurately represent the influence the law of scattering leaves to reasoned judgment the cutoff of that literature on authors. Cole and Cole (1972) used that between core and other journals. There is no clear bound- premise to show that a small number of researchers produce ary for a core collection in Bradford’s law, or in any other most scientifically important papers, and then defended the success-breeds-success model. use of citation analysis in a subsequent reply to letters to the A very common, albeit less formal expression of the editor (Cole and Cole 1974). MacRoberts and MacRoberts skewed distribution is known as the “80/20 rule”: 80% of (1989) described flaws in this assumption and the practices uses come from 20% of sources. The 80/20 rule of thumb that flow from it. Formal and informal influences are not can be applied to citations, inventories, profits, or journal always cited, authors may have bias in citing works, authors collections. In business the 80/20 rule can be expressed as “a often cite themselves, types of citation are not consistent, business gets 80 percent of its activity from 20 percent of its and citation rates vary considerably among disciplines, product line” (Rosenberg 1993, 117). The library corollary nations, and times. would be that 20% of the collection receives 80% of the use. Many citation analyses are based on data provided by One goal of this study will be to measure the degree to published citation indexes, notably the Institute for which the 80/20 rule of thumb matches patterns of profes- Scientific Information’s (ISI) Journal Citation Reports, sionals’ use of journals in communication disorders. based on data in the Science Citation Index (SCI) and Social Science Citation Index (SSCI). Critics of the reliability of data in the ISI databases have identified potential problems Problem Statement with measurement errors caused by title changes, aberrant title abbreviations, and incomplete coverage (Rice et al. A known set of core journals helps one see the boundaries 1989). Funkhouser (1996) found that a significant portion of of a discipline (Zipp 1999), and a core collection list is a valu- the references in communications journals were not covered able collection development tool. Magazines for Libraries in the ISI databases. (Katz 1969−97) identified basic periodicals considered by Citation analysis has consistently shown there to be a subject specialist librarians to be essential to library collec- highly skewed distribution of cited works, whereby suc- tions. However, contributors’ interpretations of “essential” cessful articles gain the most attention, and thus become vary a great deal, so the number of journals deemed “basic” even more successful. Success-breeds-success is often varies widely among the listed disciplines. Unfortunately, referred to as the Matthew Effect, from the book of communication disorders is not a subject heading in Matthew (13:12): “for unto every one that hath shall be Magazines for Libraries (1997), so its journals do not receive given, and he shall have abundance: but from him that a specialist’s treatment there. Neither is communication dis- 45(1) LRTS Using Citation Analysis to Pursue a Core Collection of Journals 5

orders identified as a separate subject in ISI’s Journal volumes of two journals was chosen. Purposive, or judgment, Citation Reports (Institute for Scientific Information 1996). sampling is used when probability sampling is practically The multidisciplinary nature of communication disor- impossible (Miller 1991). Probability samples are only possi- ders makes identifying a core collection difficult. The corpus ble when the population can be defined, and all members of of literature devoted exclusively to communication disorders the population can be listed. Because of the dispersed nature is small, and relatively few journals cover the entire field. As of articles used by researchers in communication disorders, a is true in many fields, many important journals are devoted probability sampling technique was practically impossible for to subspecialties, for example, the Journal of Fluency this study. Miller (1991) also states that data from judgment Disorders and Ear and Hearing. Many of the most impor- samples only suggest or indicate conclusions, meaning that tant journals for researchers in communication disorders are the purposive sampling technique used here precludes a associated with other fields, such as the Journal of the strictly defined core collection. Acoustical Society of America, Child Development, and The two journals selected were the Journal of Pediatrics. The literature researchers need thus comes from Communication Disorders and the Journal of Speech, disparate sources, including sources that are not immediate- Language, and Hearing Disorders. The Journal of Speech, ly obvious from titles, subject headings, or classification Language, and Hearing Disorders (JSLHR) was chosen numbers. because of its leadership in the field as the flagship journal A thorough search in indexes by the author unearthed of ASHA. Each issue of JSLHR has three sections, covering no published lists of core journals in communication disor- research encompassing disorders of speech, disorders of lan- ders. Therefore, there is a particular need to employ citation guage, and hearing impairments. This distribution of analysis to study the bibliometric characteristics of publica- research articles serves to cover the full field of communications serving researchers in communication disorders. This tion disorders. The Journal of Communication Disorders study will address three interrelated research questions. The (JCD), an Elsevier journal, also publishes original articles primary research question is: relating to all three areas of disorders of speech, language, and hearing. Both JSLHR and JCD mostly publish articles ■ Which journals are frequently cited by publishing based on primary research, and both are known by the researchers in communication disorders? author to be highly regarded by the faculty and frequently used by the students in communication disorders at the The method employed also provides data to answer two College of Saint Rose. Both journals state that their purpose related questions: is to address topics across the entire field, and both are specifically targeted solely to communication disorders. As ■ What proportion of citations come from various types mentioned above, many important journals focus on a sub- of sources, i.e., journals, books, conferences, tests and specialty (e.g., audiology, fluency disorders, brain injury, or assessments, and others? educational aspects of communication disorders). Many oth- ■ What is the age distribution of cited journal articles? ers cover more than communication disorders (e.g., cogni- That is, for how long do journal articles influence tive psychology, linguistics, pediatrics). The two journals authors? chosen accurately express the multidisciplinary scope of the field. Every cited work from every article in ALL issues of the Journal of Communication Disorders from 1997, 1998, and Method 1999 were entered into a spreadsheet. Data entered included the citing journal, year of the citing journal, cited work, An independent collection of data was undertaken because and year of cited work. Full journal names were entered citation analysis data published in ISI’s Journal Citation exactly as they appeared in the articles. Other sources were Reports (JCR) for sciences and social sciences do not identified by their type, such as “book,” “conference,” “dis- include communication disorders as an identified field of sertation.” The JCD for these three years yielded 2,660 cited study, and because not all significant journals are in the JCR works. Equivalent data entered from all issues of the JSLHR data set. Furthermore, citation data in the JCR on CD- from 1997 and 1999 yielded 9,034 cited works. Data were ROM are divided into separate databases for Science and not entered for the JSLHR for 1998 in order to reduce data Social Science. Journals important for communication disor- entry time, and to avoid having the data from it statistically ders are divided into two separate data sets. The separated overwhelm data from the JCD. data must be integrated to get a true picture of journal use. Only issues published since 1997 were used. It was con- From the population of journals targeted to profession- sidered important to measure, as much as possible, the cur- als in communication disorders, a purposive sample of recent rent actual use of journals. It was deemed more important 6 Black LRTS 45(1)

from a collection development standpoint to see what has column shows the number of citations, and the second shows been used recently, and less important to know what has this number as a percentage of all citations. The rightmost been used in the past. A previously reputable journal that column shows the cumulative subtotal, by percentage, of all has slipped in actual use will then receive its fair place in the citations to journals. The appendix displays data on 103 of the measurement of relative importance. Also, citation analysis 791 journals that were cited at least once. Cited sources are is unavoidably biased against new journals. Counting cita- indeed concentrated in a small number of titles, since 13% of tions from older volumes increases that bias. the cited journals provided 80% of the cited articles. Of the Since a judgment sample was employed, the degree to cited journals, 31%, or 144 additional journals, would be which this sample represents the population of communica- required to provide 90% of the cited articles. As predicted by tion disorders journals is unknown. Although the total of the theory of skewed distribution, the “tail” of infrequently works cited is 11,704, representing a substantial investment cited journals is long. Among the 791 cited journals, 156 are of data entry time, the data were drawn from only two cited only twice, and 361 are cited but once. respected sources. These results must be taken as estimates, recognizing that journals cited in all communication disorders journals vary by an unknown amount from the data reported here.

Results

Figure 1 shows the distribution of works cited by type of source (n=11,704). The books category is interpreted rather broadly, and includes reports and manuals, except for test manuals. The tests category is for all tests and assessments, and the manuals that accompany them. “Conferences” includes published proceedings, papers presented, and posters presented at conferences. The “other” category includes dissertations, theses, standards, and computer programs. Notably, of all citations, only two were to any type of Web site. Citations to journal articles comprised 7,792 of the Figure 1. Sources of Citations 11,704 works cited. Figure 2 illustrates the age distribution of cited journal articles, calculated by subtracting the year of the cited article from the year of the citing article. Articles in press were counted as being published in the year of the citing article in hand. As figure 2 shows, the age of cited articles rises quickly to those two years old, peaks at articles five years old, and gradually decreases thereafter. Sources 20 or more years old contributed 16% of all the cited articles. That fact underscores the importance of collection development as a long-term process, and indicates researchers’ need to have a reliable archival record of published research. The appendix lists, in descending order, the shortest possible list of journals that provide 80% of articles cited Figure 2. Age Distribution of Cited Articles in the sample (n=7792). The first data 45(1) LRTS Using Citation Analysis to Pursue a Core Collection of Journals 7

Title changes complicate the citation analysis. If a jour- stances must also be considered. For example, the College nal changed its name, the citation counts for the previous of Saint Rose’s program is ASHA-certified for speech-lan- title(s) were added to the current title. The one exception to guage pathology, but we have no audiology program. So the the grouping together of name changes is the journals that serials librarian and the department both believe it is unnec- preceded the Journal of Speech, Language, and Hearing essary to subscribe to the Journal of the Acoustical Society of Research. In addition to the title change for the Journal of America or Ear and Hearing, despite those journals’ high Speech [and Hearing] Disorders, that journal merged with standing in table 1. the Journal of Speech and Hearing Research in 1991. The The general stature of certain medical and scientific merged journal changed its name from Journal of Speech journals is well known, but their ranking as core titles for and Hearing Research to Journal of Speech, Language, and communication disorders came as somewhat of a surprise. Hearing Research in 1997. All versions of what is now rep- Subscriptions to Pediatrics, Science, Nature, and the New resented by Journal of Speech, Language, and Hearing England Journal of Medicine benefit researchers in this field Research contributed almost one quarter of all cited articles, more than the author would have guessed. The lesson could and 17% of articles cited in the Journal of Communication be that it is important for librarians to protect subscriptions Disorders. to general titles that serve multiple departments. Further citation analysis of communication disorders journals could provide a more broadly based and authorita- Discussion tive set of leading journals. Unfortunately, a more reliable list of core journals in communication disorders can only be The information presented here suggests a core collection in determined by expanding the sample size. This work would communication disorders, and highlights some characteristics be easier if Journal Citation Reports (Institute for Scientific of publishing within the field. Journals and books are the pri- Information 1996) recognized communication disorders as a mary sources of cited information. It is noteworthy that with- distinct discipline. As it now stands, there is no way to avoid in this sample, online journals and Web resources were not tedious and time-consuming data entry to pursue a more cited. As the theory of skewed distribution predicted, a small comprehensive citation analysis. But despite the constrained percentage of journals provided a large percentage of cited sample, this study provides objective information suggesting references. In this study, 13% of cited journals provided 80% a core of journals in communication disorders. of the citations to journal articles. Articles are used over a long period of time, with fully half of the cited articles being at Works Cited least 8 years old. The longevity of journal articles emphasizes American Speech Language Pathology Association. March 14, the need for long-term maintenance of journal collections. 2000. Highlights and trends: ASHA membership and affilia- The journals listed in table 1 suggest, but do not define, tion: Counts for mid-year 1999. Rockville, Md.: American a core collection. The selection of 80% of cited articles, in Speech Language Pathology Association. Accessed March 30, this case from journals cited 12 or more times, is an arbitrary 2000, www.asha.org/research/trends.htm. cutoff. It makes little sense to say that a journal cited 12 Bensman, S. J., and S. J. Wilder. 1998. Scientific and technical seri- times in this sample is a core journal, and one cited 11 times als holdings optimization in an inefficient market: A LSU serials redesign project exercise. Library Resources & Technical is not. But the list has to be cut off at some point. Only one Services 42: 147–242. journal was cited 12 times, five were cited 11 times, and four Bradford, S. C. 1953. Documentation. London: C. Lockwood. were cited 10 times, so there is something resembling a nat- Cole, J. R., and S. Cole. 1972. The Ortega hypothesis. Science 178: ural break at the 80% mark. Nevertheless, one should not 368–75. automatically exclude journals not listed here when collect- ———. 1974. Citation analysis. Science 183: 28–33. ing for communication disorders. The sample used was not Funkhouser, E. T. 1996. The evaluative use of citation analysis for large enough to determine that a journal not on the list is not communication journals. Human Communication Research valuable. It is also important to note that the method 22: 563–74. employed has a bias for established journals. The age of cita- Institute for Scientific Information. 1996. Journal Citation tions shown in figure 2 demonstrates that a new journal has Reports—Social sciences edition. Philadelphia: Institute for less of a chance to be frequently cited, and a great chance of Scientific Information. Katz, W. A. 1969–97. Magazines for Libraries (1st−9th ed.). New being left out of appendix A. Providence, N. J.: R. R. Bowker. Faculty opinion is very important to core collection def- Kuhn, T. H. 1996. The structure of scientific revolutions. 3d ed. inition, because faculty know their field, and give assign- Chicago: University of Chicago Press. ments that require journal use. Local use studies can MacRoberts, M. H., and B. R. MacRoberts. 1989. Problems of provide additional valuable data, since students may use dif- citation analysis: A critical review. Journal of the American ferent journals than publishing authors use. Special circum- Society of Information Science 40, no. 5: 342–49. 8 Black LRTS 45(1)

Merton, R. K. 1968. The Matthew effect in science. Science 159: Rice, R. E., et al. 1989. Journal-to-journal citation data: Issues of 56–63. validity and reliability. Scientometrics 15: 257–82. Miller, D. C. 1991. Handbook of research design and social meas- Rosenberg, J. M. 1993. Dictionary of business & management. urement. 5th ed. Newbury Park, Calif.: Sage Publications. New York: John Wiley & Sons. Oluic-Vukovic, V. 1997. Bradford’s distribution: From the classical Zipp, L. S. 1999. Core serial titles in an interdisciplinary field: The bibliometric ‘law’ to the more general stochastic models. case of environmental geology. Library Resources & Technical Journal of the American Society for Information Science 48: Services 43: 28–36. 833–42. Osareh, F. 1996. Bibliometrics, citation analysis and co-citation analysis: A review of the literature. Libri 46: 149–225.

Appendix Core Journals in Communication Disorders (n=7,792)

Frequency % of Cumulative% Cited Journal of Journal Cites Journal Cites of Journal Cites Journal of Speech and Hearing Research 1344 17.25 17.25 Journal of the Acoustical Society of America 636 8.16 25.41 Journal of Speech and Hearing Disorders and Journal of Speech Disorders 396 5.08 30.49 Brain and Language 189 2.43 32.92 Journal of Speech, Language, and Hearing Research 161 2.07 34.98 Journal of Fluency Disorders 138 1.77 36.76 Ear and Hearing 120 1.54 38.30 Language, Speech and Hearing Services in Schools 117 1.50 39.80 Journal of Communication Disorders 108 1.39 41.18 Journal of Phonetics 108 1.39 42.57 Journal of Child Language 94 1.21 43.78 Applied Psycholinguistics 93 1.19 44.97 Child Development 90 1.16 46.12 ASHA 81 1.04 47.16 American Journal of Speech-Language Pathology 76 0.98 48.14 Clinical Linguistics and Phonetics 69 0.89 49.02 Journal of Memory and Language and Journal of Verbal Learning and Verbal Behavior 64 0.82 49.85 Pediatrics 61 0.78 50.63 British Journal of Disorders of Communication and European Journal of Disorders of Communication 60 0.77 51.40 Journal of Voice 60 0.77 52.17 Science 58 0.74 52.91 Cognition 57 0.73 53.64 Seminars in Hearing 57 0.73 54.38 Psychological Review 55 0.71 55.08 Topics in Language Disorders 55 0.71 55.79 Folia Phoniatrica et Logopaedica 54 0.69 56.48 Laryngoscope 50 0.64 57.12 Journal of the American Academy of Audiology 47 0.60 57.73 Perception and Psychophysics 46 0.59 58.32 Aphasiology 44 0.56 58.88 Journal of Child Psychology and Psychiatry and Allied Disciplines 44 0.56 59.45 Phonetica 44 0.56 60.01 Language and Speech 43 0.55 60.56 Developmental Psychology 41 0.53 61.09 Neuropsychologia 41 0.53 61.61 Nature 40 0.51 62.13 Journal of Learning Disabilities 38 0.49 62.62 Journal of Neurophysiology 37 0.47 63.09 Archives of Otolaryngology Head and Neck Surgery 36 0.46 63.55 Developmental Medicine and Child Neurology 36 0.46 64.01 Electroencephalography and Clinical Neurophysiology 35 0.45 64.46 Journal of Autism and Developmental Disorders 35 0.45 64.91 Journal of Experimental Child Psychology 32 0.41 65.32 45(1) LRTS Using Citation Analysis to Pursue a Core Collection of Journals 9

Audiology 31 0.40 65.72 Journal of Experimental Psychology: Human Perception and Performance 31 0.40 66.12 Acta Otolaryngologica 29 0.37 66.49 Language 29 0.37 66.86 American Journal of Audiology 28 0.36 67.22 Hearing Research 28 0.36 67.58 Journal of Pediatrics 27 0.35 67.93 Perceptual and Motor Skills 27 0.35 68.28 Psychological Bulletin 27 0.35 68.62 Archives of Neurology 26 0.33 68.96 Brain 26 0.33 69.29 Augmentative and Alternative Communication 25 0.32 69.61 American Journal of Otology 24 0.31 69.92 Journal of Psycholinguistic Research 24 0.31 70.23 Annals of Otology, Rhinology and Laryngology 23 0.30 70.52 Cortex 23 0.30 70.82 British Journal of Audiology 22 0.28 71.10 Experimental Brain Research 22 0.28 71.38 First Language 21 0.27 71.65 Neurology 21 0.27 71.92 Quarterly Journal of Experimental Psychology (A and B) 21 0.27 72.19 Scandinavian Audiology 21 0.27 72.46 Language Acquisition 20 0.26 72.72 New England Journal of Medicine 20 0.26 72.97 Speech Monographs and Communication Monographs 20 0.26 73.23 American Journal of Public Health 19 0.24 73.47 Annals of Neurology 19 0.24 73.72 Journal of Educational Psychology 19 0.24 73.96 Seminars in Speech and Language 19 0.24 74.20 Volta Review 19 0.24 74.45 Clinical Aphasiology 18 0.23 74.68 Cognitive Neuropsychology 18 0.23 74.91 Journal of Experimental Psychology: Learning, Memory and Cognition 18 0.23 75.14 Journal of Applied Physiology 17 0.22 75.36 Memory and Cognition 17 0.22 75.58 Annals of Dyslexia 16 0.21 75.78 Cognitive Psychology 16 0.21 75.99 Journal of Applied Behavior Analysis 16 0.21 76.19 Journal of the Academy of Rehabilitative Audiology 16 0.21 76.40 Linguistic Inquiry 16 0.21 76.60 Otolaryngology Head and Neck Surgery 16 0.21 76.81 Speech Communication 16 0.21 77.01 Brain Research 15 0.19 77.21 Merrill-Palmer Quarterly 15 0.19 77.40 Monographs of the Society for Research in Child Development 15 0.19 77.59 American Journal of Medical Genetics 14 0.18 77.77 Audiology Today 14 0.18 77.95 Cleft Palate-Craniofacial Journal 14 0.18 78.13 International Journal of Pediatric Otorhinolaryngology 14 0.18 78.31 Journal of Gerontology (split into A and B) 14 0.18 78.49 Journal of Neurology, Neurosurgery, and Psychiatry 14 0.18 78.67 American Annals of the Deaf 13 0.17 78.84 Dysphagia 13 0.17 79.00 Infant Behavior and Development 13 0.17 79.17 JAMA 13 0.17 79.34 Journal of the American Academy of Child and Adolescent Psychiatry 13 0.17 79.50 Psychology and Aging 13 0.17 79.67 Sign Language Studies 13 0.17 79.84 British Journal of Psychology 12 0.15 79.99 10 LRTS 45(1)

Book Vendor Records in the OCLC Database Boon or Bane?

Laura D. Shedenhelm and Bartley A. Burk

This case study is based on a 1998 sample of recently acquired Spanish-language firm-ordered materials, all of which had vendor records in the OCLC database. Vendor records were compared to final fully cataloged records to study differences in the basic bibliographic description fields (1xx, 245, 300, 4xx, 5xx). Identified were the types of errors found in the records and the duplication rate with records already in the database (full LC and member records, partial member records, and other vendors). Both areas are problematic for cataloging units. Secondary research objectives included tracking titles for usable copy cataloging and analyzing the cost impact for typical cataloging workflow. The researchers conclude that the records, though sometimes problematic, are useful. Suggestions are given for areas of improvement.

n recent years, there has been an exceptionally close scrutiny over public Ifunds. Libraries are included in this realm and are among the most cost-con- scious entities in the public sector. This undoubtedly reflects our innately con- servative approach to fulfill our responsibility to acquire, maintain, and preserve the intellectual and cultural patrimony of our parent institutions and to serve our clients. While we have watched the steady increase of prices for the materials we purchase, we have sought innovative ways to hold down costs for acquiring and providing access to our collections. Book vendors are also seeking ways to lower their overhead costs, provide needed services to their clients, acquire new clients, and maintain a profit margin which will allow them to stay in business. There is a symbiotic relationship between libraries and book vendors that is nec- Laura D. Shedenhelm is Special essary for both groups to succeed in a fast-paced and competitive world. This Projects Cataloger, University of study looks at part of the symbiosis that appears to be deleterious to one part of Georgia, Main Library, Athens, and that relationship: the impact on cataloging processing units within libraries of Bartley A. Burk is Social Sciences Cataloger, University of Notre Dame, vendor-produced records in the national bibliographic databases. Indiana. Vendor records are very brief bibliographic records originally designed to Terry Peet of the Library of Congress advertise an item for sale by the vendor. They are based on the files the vendors (LC) kindly provided the authors with use to create their own sales catalogs and include minimal information: the an oral history of the LC’s use of vendor records. Russell Marr, also from LC, author and title of the work, publication information, and extent of the item. helped gather information about ven- Often, they also will have notes about the work, “edition” information, and some dors’ use of LC’s authority files. Jay subject analysis. When this information is accurate, it is very useful for bibliog- Weitz from OCLC sent information about OCLC’s deduplication process. raphers who must decide if they want to purchase the item for their collections. The authors also wish to thank Richard However, a problem arises when the data presented do not accurately represent Shedenhelm, John Riemer, and the bibliographic item they are meant to describe. The descriptions in vendor Barbara Winters, all from the University of Georgia, for their helpful comments catalogs are often so brief it takes real skill to decide if the item would be a during the drafting of this article. unique addition to a collection. For example, “edition” entries in these catalogs Manuscript received January 27, 2000; often merely indicate new printings, not new versions of the works. The form of accepted May 15, 2000. 45(1) LRTS Book Vendor Records in the OCLC Database 11

the author’s name, and sometimes even the choice of author were any savings as a result of vendor records being entry, rarely matches a library’s entry in its catalog. Once the in the national databases. item is purchased for a collection, catalogers must describe it in terms of library cataloging rules more detailed than Spanish-language catalogers were among the first to feel those adequate to advertise an item. Catalogers must distin- the impact of vendor records. During the last few years, guish the basic bibliographic unit from all other renditions though, other non-English-language catalogers have begun to of the intellectual content presented, and they must make it see similar records appear for their materials. Most European fit intelligently into the larger bibliographic universe. What vendors now routinely send tapes to be loaded into the bibli- happens, then, when these minimal records are added to the ographic utilities. While this study looks specifically at records national databases that traditionally provide accurate and from Spanish-language vendors, the basic issues are applica- full cataloging that generally meet industrywide agreed- ble to all of these records. This was verified through col- upon standards? Is there an impact on the work performed leagues working with German-language material at University in libraries’ cataloging units? of Notre Dame and French-language materials at University Our premise is that there is an impact. We took a small of Georgia. We note here that these issues apply most direct- sample of vendor records as a basis for a study on their ly to OCLC, which uses the single “master record” method impact on a typical cataloging workflow. We looked at: for displaying bibliographic cataloging and holdings. Our colleagues using the RLIN utility with its “clustering” technique 1. How much manipulation of the data in vendor-pro- have fewer dilemmas with these records. duced records must be done in order to bring them into alignment with library standards for cataloging. We did this by comparing the original versions of the History of Spanish-language Vendor vendor records to the title pages of the works they Records in Bibliographic Utilities represented. We asked: Are there errors in vendor records that make them difficult to utilize as nation- The history of vendor records in the bibliographic utilities ally acceptable standardized library cataloging goes back to the early 1990s (Peet 1999). During 1990 and records? If the answer is yes, is there a pattern to the 1991, representatives of Casalini Libri approached the errors? Library of Congress with the idea of using electronically for- 2. How well vendor records fit into the greater biblio- matted vendor records to streamline the acquisition process. graphic universe. We searched the OCLC database Unfortunately, the idea was “before its time,” since the for any other records that would also represent the online system at LC was not configured to be able to take items in the sample. We asked: Do vendor records advantage of this technological enhancement. LC decided duplicate any records in the national bibliographic not to pursue this idea. databases? If so, at what rate? Various factors converged during the next few years forcing the Library of Congress to take a more aggressive These two areas, editing of records and determining approach to implementing computer-based processing. appropriate records to use for cataloging, constitute the There was a federal hiring freeze and a vacancy rate of majority of work performed in cataloging units. Anything library technicians at LC that reached nearly 40.00%. In that negatively impacts these two areas would have an unfa- May 1993, at the Guadalajara meeting of SALALM vorable effect on cataloging production. Timely and cost- (Seminar on the Acquisition of Latin American Library effective cataloging is a primary goal for any cataloging unit. Materials), members of LC’s Hispanic Acquisitions Because of this we realized that there were two other ques- approached Puvill Libros with this same idea of using elec- tions (which we will call 3 and 4) that our study needed to tronically formatted vendor records in order to address the address: growing backlog of materials at the library that needed to move through the acquisition process. By 1993, LC had an 3. How soon does full-level copy cataloging appear in overall backlog of 8,500 recent Spanish imprints (two- to the national bibliographic databases for items repre- three-year-old publications) that needed minimal-level sented by vendor records? We tracked the sample records in their database. After several months of testing, items through an eleven-month cycle to see when any LC received the first set of MARC-formatted records from source of copy cataloging would appear. We per- Puvill in early July 1994. Always open to pursuing good, formed a follow-up search fourteen months later. apparently cost-saving ideas, Puvill approached OCLC in 4. Do vendor records provide a cost-efficient method 1995, suggesting the sharing of these records with a wider for giving access to our collections? We looked at audience. Many other vendors have followed this same path, costs for a typical cataloging routine to see if there leading us to the current situation. 12 Shedenhelm and Burk LRTS 45(1)

Literature Review and Fox review the theory that minimal-level records in the database help librarians predict when full-level records will A quick search of library literature will show that minimal- appear. They use a similar technique to our study, checking level records in the utilities is not a new idea. Our initial lit- for record upgrades over a six-month period. They found erature search indicated that there had been nothing written that minimal-level records duplicated full-cataloging mem- about vendor-produced bibliographic records. However, ber records at a rate of 20.00%. They also indicate that cat- this has recently changed. alogers discovered upgrading minimal-level records took as much time to catalog as items without copy, and required Vendor-produced Bibliographic Records additional work to verify if there were duplicate records in the OCLC database. When duplicates were found, OCLC Beall “subjectively examines the impact the addition of ven- was notified, adding more time to the process of what dor record to the bibliographic utilities, chiefly OCLC, has should be inexpensive, simple copy cataloging. They cite a had on cataloging” (Beall 2000, 230). He analyzed vendor 1990 survey of Association of Research Libraries as sup- records for content and quality, then looked at the impact on porting their observations. Similar discussions on upgrading cataloging and patron access. His study is based on opinion minimal-level records can be found in case studies by and calls for three areas of quantitative analysis. Our study Ferguson (1991) and Handman (1991). validates much of his commentary, and addresses in part his Horny (1991) and Intner (1994) both look at the prom- suggestion for research about the quantity of vendor records ise of minimal-level cataloging versus the end product. that were enhanced after one to two years in the utility. Horny points out that if the materials are not processed through the system, they are not available to the patron. Minimal-level Records and Duplication Minimal-level cataloging promises to move items through more quickly. However, she notes that the information most Two of the best histories about minimal-level records are by easily left out of records is the data that is most easily and Patton (1991) and Stamm (1996). Everyone is aware of CIP quickly entered into them. The most time-consuming aspect records and those of us who have been around for a while of cataloging is authority control where “savings are difficult will remember the early days of mass tape-loading projects to achieve” (Horny 1991, 10). Intner concludes that the that added many minimal acquisition-level records to items that would most benefit from minimal-level cataloging OCLC. In our drive to get the most out of our processing are those most easily identified and available rather than the dollars, we seem to be wedded to the idea that “something most esoteric (Intner 1994). is better than nothing.” Yet, as professionals, we have questioned that marriage Duplication and Errors often through the years. Fox and Preece (1991) and Preece and Fox (1992) cover issues most closely related to vendor Beyond Fox and Preece, much has been written about the records. They discuss three reasons to upgrade minimal- difficulties arising in all areas of library work due to duplica- level records: (1) to provide increased patron service tion of records in the database. Johnson and Josel (1981) dis- through complete cataloging; (2) to maintain the library’s cussed the types of errors that cause duplicates and the commitment to quality cataloging; and (3) to enhance resulting costs of duplication. Wanninger (1982) looked at record quality for the online catalog (Fox and Preece 1991). the impact of duplication on search results and difficulties As with vendor records, the minimal-level records in their related to multiple searching strategies. O’Neill, Rogers, and study had no subject headings and there was no authority Oskins (1993) looked at typical errors found in records that control for access points. They note that accuracy in tran- result in duplication in the OCLC database. The character- scription and coding of the title and extent fields is an area istics include “(1) typographical errors, (2) erroneous tags of highest concern: “[to] support key word title access, accu- and subfield codes, (3) omitted information, and (4) incon- rate transcription of the entire 245 field is critical” (Fox and sistencies between the variable and fixed fields” (O’Neill, Preece 1991, 29). They conclude that very experienced cat- Rogers, and Oskins 1993, 61). Our findings closely parallel alogers must do a detailed review of the entire record. They theirs. They assert that it is transcription in a form that is go on to recommend ensuring that the minimal-level record similar but not exactly the same that leads to most duplicate is not a duplicate in the OCLC database. “Duplicate” means records, and point to batch loading as a major source of one or more record(s) representing one bibliographic item duplicate records. The most significant characteristics iden- that do not exactly match, resulting in a duplication of the tified for duplication occurred in the date, author, or pub- OCLC master record, which will not automatically combine lisher areas of the record. In particular, one-third of the with the master record when OCLC’s merge program is run. duplicated records contained author entries that did not In their subsequent article (Preece and Fox 1992), Preece match. It should not be assumed, however, that only records 45(1) LRTS Book Vendor Records in the OCLC Database 13

that are duplicated show these characteristics. A study con- the record just as it originally appears from the vendor. ducted by Romero (1994) that looked at typical errors made Additionally, name headings would have to be verified, as it by both entry-level and experienced catalogers indicated was unlikely that the vendors would create the nationally that there were areas of difficulty in cataloging in general. compliant access points. We also wanted to see how these She found that most errors occurred in choice of heading records fit into the overall picture of bibliographic copy and description, but that typographical errors were minimal. available to us in our daily work. Did these records represent In the area of description, the publication field was the most any duplication of records already in the bibliographic utili- problematic. Our findings corroborate this. Also, the results ties? If the answer is yes, then how much duplication is cre- of our study will demonstrate that all of the problems seen ated? Finally, we realized that the deciding factor about the with earlier versions of batch-loading minimal-level records utility of these records would be based on whether they are intrinsic in loading of vendor tapes into the utilities. were cost-efficient or not. At what point could these records be handled by the least costly staff? A précis of the methodology and preliminary data used here was presented at the History of the Project meeting of the Cataloging and Bibliographic Subcommittee during the 1999 annual meeting of SALALM held in From the first appearance of vendor records in the biblio- Nashville, Tennessee. graphic utilities, there has been intense, often heated, discussion about these records during the SALALM annual meetings. While it was noted that acquisitions departments, Method including the Library of Congress’ Hispanic Acquisitions Section, find these records useful, there was a diversity of The purpose of this small study is to see if the complaints opinion among catalogers at SALALM about their overall about the quality of vendor records (which were raised dur- utility. The SALALM Cataloging and Bibliographic ing the SALALM meetings), the rate of duplication in the Technology Subcommittee has members from a variety of OCLC database, and the resultant increased cataloging public and private, large and small institutions that subscribe workload and cost are justified. To do this, we compared the to either RLIN or OCLC, or both. As such, the subcommit- vendor record to the item, seeking any errors in transcrip- tee is representative of the greater library community. tion or choice of data transcribed; we searched the OCLC Variations in processing workflow and attitudes about patron database for any duplicate records; we tracked the records access at different libraries influence whether these records for potential use in copy cataloging; and we analyzed the are more or less troublesome. This is especially true when costs. coupled with a library’s decision to catalog the items locally Our sample was gathered from the workflow at the or nationally. SALALM members that catalog locally and University of Notre Dame Hesburgh Library, which uses only attach their holdings to a national utility through tape- vendor records in its acquisition process, but is not an loads have fewer complaints about vendor records than do OCLC enhance library. All of the titles would have been members who catalog directly onto the utilities. searched in OCLC at the point of order. If a record existed The seemingly endless discussion about these records at that time, it would have been downloaded into the local prompted some of us at SALALM to do a study of the ven- system to generate the order. Otherwise, a record would dor records. Our hypothesis was that, while possibly useful have been created on the local system. The items were at the acquisition stage, these records posed several prob- always searched by the International Standard Book lems for catalogers. Chief among these problems is that con- Number (ISBN). This is a simple and inexpensive search verting these records to nationally acceptable standardized that should give good results provided the ISBN appears in library cataloging records would be time-consuming. We felt the record. The assumption was that this search takes that it was easier and more cost-effective to create an origi- approximately five minutes per item. nal cataloging record. We believed that comparing the When the items were received, the local database was unenhanced vendor record to the actual item would illus- searched. Items with full- or nearly full-cataloging records trate this. We knew that the vendor records would always were appropriately distributed to the copy-cataloging unit. lack specific information we require in our fully cataloged Items with strictly acquisition-level records were researched records: classification (in this case Library of Congress by ISBN in the OCLC database, then distributed accord- Classification, or LCC), bibliographic and content notes in ingly. Normally, titles lacking full cataloging are put into a English, and subject access through Library of Congress searching cycle for a maximum of 18 months. About every Subject Headings (LCSH). All of this must be added to the three months the title is searched for an acceptable record, record before it can pass out of processing as fully usable and sent through processing or put back into the search cataloging, which means that the library can never accept cycle until it completes 18 months in the cycle. If there is no 14 Shedenhelm and Burk LRTS 45(1)

full cataloging at the end of that period, the item is sent to databases; (2) the amount of this duplication; (3) timeliness the professional staff for local processing. of copy cataloging for items represented by vendor records; We collected a set of 64 titles from all items routed to and (4) processing costs for cataloging items using vendor- the social sciences cataloger at Notre Dame from two ship- generated records. ments from Puvill Libros received in May and June of 1998. The sample was of titles that had only vendor records avail- Errors able when received. The original versions of the vendor records associated with the sample titles were printed. We As noted in the literature review, O’Neill, Rogers, and Oskins kept these as examples of how they appeared in OCLC, (1993) and Romero (1994) found characteristic types of errors prior to any upgrade or merging with other records. records in the database. These errors are consistent with our In early July 1998, the title pages and versos of each findings for vendor records (see tables 1 and 2). While we saw sample title were photocopied and matched with the origi- some problems with accurate transcription, we believe the nal vendor record. Using this process, we sometimes missed typing errors were typical of any randomly chosen set of series information that would have appeared on covers, records from the database. There was also the occasional odd series title pages, or spines. Therefore we will not include decision for 245 filing indicators. These two problem areas any discussion of the series fields in our analysis. The title would result in difficulty locating the record in the database page and verso information was then compared to the orig- and could contribute to duplication, but overall were minor. inal versions of the OCLC records, noting any discrepan- Romero found that both experienced and beginning cies, such as typographical errors, omissions of data required catalogers made the most errors (41.60% and 49.06%, by the cataloging code, miscoding of fields, etc. The 300 respectively) in (1) choice and form of headings and (2) field (pagination, etc.), in general, was omitted from this description of the items (Romero 1994). We also found that review as the work was being done without the book. The the most difficulties were in these two areas (see tables 1 exception was for multivolume sets. The issues related to and 2). The 260 field was particularly problematic due to these are discussed under “duplication” below. inaccurate presentation of the publisher (33 of 49 errors, or During late July and early August of 1998, the OCLC 67.35% of description errors). From the information in table database was searched for all records that might match the 1, one can conclude that the error rate in the description in title pages. A combination of techniques was used in this vendor records is significantly higher (50.00%) than in process. We searched by ISBN, all types of derived search- records produced by people with minimal professional es, and scan title in order to ensure that we found any record training. Additionally, one-fifth of the records (20.31%) had that could be used in the cataloging process. This process errors in choice or form of headings, either main or added was repeated in October of 1998 and April of 1999. Due to entries. the appearance of the Beall (2000) article, another search Accurately editing fields that are inconsistently entered was done in July of 2000 for any title that had not received is time-consuming and difficult. Having to spend extra time full cataloging by April of 1999. Each time the records were meticulously correcting a record automatically raises the searched, detailed notes were made about the records cost of using it. Romero concluded that these were errors found. These notes include the number and type of dupli- that could only be addressed through improved training. We cate records found (e.g., other records from LC, a member library, or a vendor), the number of holding libraries for each record, differences in the records versus the title pages, Table 1. Types of Errors in Vendor Records and differences in the records that might cause them not to merge using a duplicate-detection algorithm. Choice/ Bibliographic Finally, in order to try to analyze costs, we kept track of Typographical Form of Headings Description when each item would have been able to move through the # of % of # of % of # of % of records records records records records records copy-cataloging process that requires no additions or changes, 7 10.94 13 20.31 49 76.56 the least expensive processing. Then, we looked at the overall costs associated with the processing of these titles.

Table 2. Multiple Error Types per Vendor Record

Analyzing Collected Data Typing and Typing and Heading and Typing, Heading, Heading Description Description and Description Our data involves four areas: (1) errors in the vendor # of % of # of % of # of % of # of % of records records records records records records records records records that lead to increased editing time, and that 1 1.56 4 6.25 7 10.94 1 1.56 may be the cause of duplication in the bibliographic 45(1) LRTS Book Vendor Records in the OCLC Database 15

absolutely agree with this assessment: these problems could loging to 25%. By the third search, the duplication rate had be overcome through training programs developed specifi- risen to 37% of the total. Of the 24 titles duplicated, 8 dupli- cally for vendors. cates had been merged. Also, 23 had either been upgraded or With regard to form of headings, Russell Marr, Library the new duplicate represented full cataloging. The other title of Congress, Hispanic Acquisitions Section, helped us had been upgraded but lacked LC classification. research the issue of vendors’ access to the LC name author- We believe that the two most common categories of ity file. According to LC personnel, this file is not used by errors noted above, description errors and incorrect form or the vendors that LC contracts for original acquisition tapes choice of headings, are the roots of the duplication found in (Marr 1998). They would need to contract with OCLC or our sample. This is due to the very broad de-duplication RLIN for access to it. The use of authorized versions of per- algorithm used by OCLC that results in a lack of match sonal and corporate names is the heart of appropriate access when the program is run. Jay Weitz, Consulting Database and collocation in our catalogs. Anyone who has ever Specialist for OCLC, sent us information about the attempted to look for information in periodical and newspa- Duplicate Detection and Resolution (DDR) software that per indexes about Muammar Qaddafi can attest to the prob- was developed about a decade ago. He stated that the lem inherent with multiple ways of entering a name! While “DDR runs through the WorldCat database on an irregular most of the personal names in the sample could be relative- basis (currently, roughly every six months)” (Weitz 2000a). ly easily identified and corrected, corporate names, espe- He indicated that the algorithm uses fourteen descriptive cially conference names, were rarely correct. These were elements on which to form a match for merging records. usually tagged as 100 fields instead of the appropriate 110 or Also, there are about ten conditions that prevent merges. 111 designation. Again, the correct tagging of the fields He noted that more than one million duplicate records for could be easily improved with minimal training, but signifi- the books format have been merged since DDR first ran in cantly more in-depth instruction would be necessary for June 1991. In his subsequent message, he listed the ele- confident construction of corporate names to fit into the ments included in the algorithm: “cataloging library; LCCN bibliographic setting that our libraries use. The lack of [Library of Congress Control Number]; ISBN; government authorized headings in the vendor records has the potential document classification number; media; author; title; state- to be very costly over time. When one considers that the ment of responsibility; edition statement; place of publica- correction to name fields must be done by every library that tion and publisher; publication date; number of pages or uses the record, until such time as a library locks and cor- volumes; size; and series statement” (Weitz 2000b). He went rects the record, one begins to see the extent of the associ- on to say “we tend to err on the side of adding or leaving ated costs. Unfortunately, without daily searching of each duplicates rather than merging away unique records when- record, we would have no way of knowing at what exact ever there is uncertainty” (Weitz 2000b). Given the need to point any particular record was upgraded, or how many match on publication data and choice and form of author, libraries had used it prior to the enhancement. Therefore, the areas with the most errors in the vendor records, it is not we cannot state unequivocally how much libraries through- surprising that only eight of the duplicate records found out the world are expending toward these corrections. during our study had been merged by the time of our third search of the database. It is also not surprising that so many Duplication duplicates are created at the initial tape loads. It would be interesting to know, of the eight merged records, how many Duplication of records in the OCLC database is problemat- were merged automatically and how many were reported by ic for efficient cataloging. We found that vendor records diligent libraries trying to clean the database. exacerbated this problem. Our initial search, using the vari- Whenever there is duplication in the database, the cost ety of techniques described above, yielded duplicate records of locating appropriate copy cataloging automatically increas- for 17 of the titles, 12 of which had usable cataloging copy. es. Instead of being able to do a simple numeric search of an This means that 26.56% of our titles were duplicated in the ISBN, which should result in one record in OCLC, the database. Also, 18.75% of our sample had usable copy in the database without the distraction of the vendor records. (See table 3.) The second and third search cycles increased the Table 3. Duplicates duplication rate to 31.25% and 37.5%, respectively. During the second search, one of the original records was # of Duplicate % of # of Records % of upgraded by the Library of Congress (LC), one was upgrad- Date Records Total with Full Copy Total ed by a member library, but also duplicated by a LC record, Aug. 1998 17 26.6 12 18.8 Oct. 1998 20 31.3 16 25.0 and two were duplicated (one by LC and one by a member Apr. 1999 24 37.5 23 35.9 record). This raised the overall rate of available copy cata- 16 Shedenhelm and Burk LRTS 45(1)

library must pay more experienced, and therefore more 12 titles that did not have full-level cataloging at the end of expensive, personnel to distinguish among several records, our initial project. We hoped our findings would shed light and decide which is the best choice. Beall (2000) also shows on this issue. Our findings do not directly answer Beall’s that libraries, which had previously been able to utilize question since we looked for any available copy, not just OCLC’s Cataloging Micro Enhancer (CatME) software to enhancements to the vendor records. However, our findings great advantage in finding suitable copy cataloging, are now show that full-level cataloging is available for items repre- stymied by the influx of minimal-level records that the sented by vendor-produced records within his timeframe. In CatME software cannot distinguish from full-level cata- table 4, we show that 61 of the 64 sample titles had full cataloging. The automated search results in multiple records loging at the end of two years. This accounted for 95.31% of being downloaded, and the consequent increase in cost. The our sample. We agree that further research needs to be done issue of duplication for vendor-produced records is particu- in this area. In particular, more study needs to be done based larly significant for Spanish-language cataloging as there are on the criticism that less original cataloging is being done multiple vendors giving records to OCLC. Our study showed because of the shift of the workload to upgrade these non- several cases when there were multiple vendor records for a standard minimal-level records (Beall 2000). single title. Another issue raised by Beall concerns uploading The issue of multiple vendor records for a single title improved versions of the vendor records for national con- was particularly evident with multi-volume sets. In our sam- sumption (Beall 2000). An earlier study by Sercan (1994) ple, we found separate records for individual volumes of found a marked decrease in the amount of full cataloging multivolume sets. Standard cataloging practice dictates that for Spanish-language materials available between 1983 and these volumes would be described on one bibliographic 1992 (18.00% versus 4.00% for LC copy, and 13.00% ver- record in most cases. The exception, of course, would be a sus 5.00% for RLIN member copy). Erbolato-Ramsey and monographic set that warranted analysis of each individual Grover’s (1994) findings “would seem to indicate that most volume (a rare case). We had two cases of multivolume/mul- libraries accept and input less than full AACR2r [Anglo- tirecords for items that were fully cataloged incomplete American Cataloging Rules, 2d ed., rev.] level records monographic sets. OCLC’s reliance on machine-driven online and do not update them [nationally] at a later time” merge algorithms to identify and merge duplicate records (83). Beall points to the lack of financial incentive as a like- will not work in these cases. We must rely on people to report ly reason for the decrease in upgraded copy (Beall 2000). these for manual merging, thereby increasing the cost of hav- Our findings do not seem to support this theory. However, ing them in the database (Johnson and Josel 1981). our sample is extremely small given the overall output of Spanish-language publications. A larger study would be Tracking Copy useful.

When we looked at the time frame for appearance of copy Costs cataloging for the titles in our sample, we were pleasantly surprised. There was a steady increase in the amount of copy The costs described here are based on the 1998 INCOLSA available through either upgrades of the vendor record or network prices and the labor costs at Notre Dame for that duplication of these records by full-level LC or member- year. The following were the labor costs: support staff for generated records (see table 4). Slightly more than 20.00% transcription = $0.33 per item (figured at $15.00 per hour of the titles had usable copy at the time of the first search and 5 minutes per item); professional completion of records period during the summer of 1998. Twelve of the 13, how- lacking only LCC = $3.75 per item; professional completion ever, came from duplicate records. In only one case was the for other records = $21.86 per item. Network costs were: record an upgrade of the vendor record, and it was done by numeric or derived search = $0.34; update = $0.43; export Notre Dame. By October of 1998, almost half (46.87%) had (to local system) = $0.12. Network credits are: nonenhance usable copy. When the last search was done in April of 1999, 11 months after the sample records were gathered, 81% had usable copy. For our original study period, only 12 remained without complete cataloging. Of those, 2 only needed a clas- Table 4. Tracking Copy sification number to complete the cataloging. # with % with # lacking % lacking Beall posed three research questions, one of which Date full copy full copy copy copy focused on percentage of vendor records enhanced at the Aug. 1998 13 20.31 51 79.69 end of a one-year period, and a two-year period (Beall 2000). Oct. 1998 30 46.87 34 53.13 Apr. 1999 52 81.25 12 18.75 Since we had been looking at availability of full cataloging for July 2000 61 95.31 3 4.69 titles represented by vendor records, we did searches for the 45(1) LRTS Book Vendor Records in the OCLC Database 17

upgrade = $2.10; enrichment (call numbers, subject head- nationally acceptable standardized library cataloging ings, etc.) = $0.52; nonenhance original addition = $3.83. records? Is this reflected in errors in vendor records Beyond these specific costs, one must remember that there that make it difficult to utilize them as cataloging are additional hidden costs for sorting through multiple records? records as identified in the section about duplication above. 2. What level of duplication in the bibliographic data- Table 5 shows the per-item processing costs during the bases do vendor records represent? original one-year period for titles in our sample using the 3. When does copy cataloging for titles represented by cycle of searching scenario earlier. Overall, this information vendor records appear in bibliographic databases? is fairly irrelevant since the items would have been 4. What are the costs associated with vendor records in processed regardless of the cost. However, these costs show the database? the same disincentive to upgrade the record that Beall found (Beall 2000): an enrichment credit of $0.52 to add the First, we determined that the vendor records would classification number to a record does not compensate for never fulfill the requirements for library cataloging records the processing costs of $6.87. Similarly, a processing cost of since they will always lack classification and subject headings $24.98 for an item needing essentially full cataloging will needed for our catalogs. We also found that there was a pat- now only gain a credit of $2.10 to upgrade nationally where tern of data-entry errors that placed a burden on the library it formerly would have been credited at $3.83. This cost community to correct. Accurate editing is a time-consuming structure would seem to undermine the tenet of cooperation and labor-intensive process in the best of settings. The types that traditionally has been the foundation of OCLC. We of errors found in the vendor records were not usually simple begin to see the dilemma presented by Sercan (1994) and typographical ones. Rather, they were content-oriented and Erbolato-Ramsey and Grover (1994). Is there enough finan- inconsistent, thus tended to increase the time necessary to cial incentive to upgrade these records, or will we begin to review each record. Among the content errors was choice and see less full-level cataloging at the national level as budgets form of headings. We concur with Romero (1994) that appro- for technical process continue to tighten? priate training, in this case for vendors wanting their acquisition-level records represented in the OCLC database, would Conclusions help alleviate many of these problems. Further we believe that access through OCLC to the name authority files would Our original questions were: improve the quality of the headings on the vendor records. Second, we found that these errors led to duplication in 1. How much manipulation of the records must be done the database. Of our sample, 37.50% was duplicated by full- for vendor records to fulfill the requirements of the level records already in the database or added during our

Table 5. Processing Costs in Dollars per Item

Pre-order Search Total or Full Cataloging Available Search Export Labor Subtotal at Receipt Update Labor Sub-total On receipt 0.335 0.12 0.33 0.785 0.335 0.43 0.33 1.88 1st cycle 0.335 0.12 0.33 0.785 0.335 0.33 1.45 2nd cycle 0.335 0.12 0.33 0.785 0.335 0.33 1.45 3rd cycle 0.335 0.12 0.33 0.785 0.335 0.33 1.45 3rd cycle plus class no 0.335 0.12 0.33 0.785 0.335 0.33 1.45 3rd cycle plus full review 0.335 0.12 0.33 0.785 0.335 0.33 1.45

1st Cycle 2nd Cycle Total or Full Cataloging Available Search Update Labor Subtotal Search Update Labor Subtotal 1st cycle 0.335 0.43 0.33 2.55 2nd cycle 0.335 0.33 2.12 0.335 0.43 0.33 3.21 3rd cycle 0.335 0.33 2.12 0.335 0.33 2.78 3rd cycle plus class no 0.335 0.33 2.12 0.335 0.33 2.78 3rd cycle plus full review 0.335 0.33 2.12 0.335 0.33 2.78

3rd Cycle Full Cataloging Available Search Update Labor Total Cost 3rd cycle 0.335 0.43 0.33 3.88 3rd cycle plus class no 0.335 3.75 6.88 3rd cycle plus full review 0.335 21.86 24.98 18 Shedenhelm and Burk LRTS 45(1)

study. Therefore, more experienced people would have been Recommendations needed during the searching process to accurately identify appropriate cataloging copy. We determined that OCLC’s Vendor records will be a permanent part of the bibliograph- duplicate-detection algorithm errs on the side of adding ic universe for processing materials in our libraries because duplicates in order to ensure unique records are not inap- they provide early notification of new materials. However, propriately merged (Weitz 2000a). This means that we must beyond timeliness, high quality records are also important rely on people to report duplication as it is found. However, for both the national databases and our local library catalogs. Johnson and Josel (1981) concluded that this is a course of We believe that by adopting some or all of the following rec- action too expensive to pursue consistently. ommendations the bibliographic quality of vendor records The corollary to the duplication issue, however, is that will improve and the level of acceptance in the library com- two-thirds of the titles in our sample were new to the data- munity will increase. base. It is valid to ask, then, would there have been more or Recommendation 1: OCLC should initiate a rigorous less full-level cataloging in the database if the vendor training campaign for all vendors adding records to the bib- records did not exist? Does having a base record from which liographic utilities. This training should concentrate on the to work help or hinder the production of final cataloging? It bibliographic description standards used by the majority of can be argued that, since materials go through the acquisi- OCLC’s constituency. tion process more quickly, they are in the cataloging work- OCLC expects members to adhere to national-level flow more quickly, and therefore are attended to sooner standards for the records we place into the database. It is than before vendor records were available. Further study reasonable for members to expect this quality for any record would have to be done in this area to verify any speed of we find in the database. Sally McCallum, Chief of LC’s access or costs savings this might represent. Network Development and MARC Standards Office would Next we looked at when full cataloging appeared for titles also seem to support this idea: represented by vendor records. We could not determine full- level cataloging availability based solely on the upgrade of [I]t is generally the responsibility of sending organ- vendor records because often full cataloging came from izations to make records conform to community records that duplicated the vendor record. However, by track- developed and approved standards, and if the orig- ing the materials through a cycle of searching the OCLC data- inal sender does not do this then many to whom the base for cataloging copy, we were able to determine how original sender distributed have to take on duplica- quickly they would have been made fully accessible to the tive work (McCallum 1997). library’s patrons. We determined that 81.25% of the titles had full cataloging at the end of the original study. In response to The authors sent examples of typical errors found in this Beall (2000), we searched the titles that lacked full cataloging study to the Library of Congress for a training program it at the end of the original study. This revealed that a total of developed in the fall of 1998. These examples could be the 95.31% of our sample had full-level cataloging available at the basis for a training program to refine all vendor records to end of two years: a very good return. meet the community-developed standards noted above. Finally, we looked at processing costs. Do vendor Recommendation 2: We encourage OCLC to make records provide a cost-efficient technique for providing the authority files available to vendors. Further, we encour- quick access to our collections? Our answer would be a qual- age extensive training in the construction and application of ified yes. If our 95% rate of full cataloging over a two-year these files and the headings they contain. period is representative of vendor records in the database, One-fifth of the errors in our sample were from incor- and as long as a handful of libraries are willing to continue rect headings. Having the correct form of the name will to bear the full burden to upgrade these records, the rest of lower the cost of upgrading the record. It will also help in us will enjoy inexpensive cataloging. However, should the the deduplication process given that “author” is one of the burden become too great and the reward remain so little, we fourteen elements used in the OCLC deduplication pro- could well begin to see a drop in the percentage of full-level gram (Weitz 2000b). cataloging available. Further study would need to be done to Recommendation 3: We strongly encourage initiation determine if there is any other positive impact from the of financial credit to libraries reporting duplicates in the presence of vendor records. In particular, more analysis of OCLC database. the impact on the costs of processing due to the duplication Every library must spend precious staffing funds wading rate associated with vendor records would be useful. Also, a through multiple records to find full cataloging. Libraries also study needs to be done to see if there is any correlation spend funds notifying OCLC of these duplicates. between the addition of new and unique titles entered as Consequently, this process is rarely done according to Johnson vendor records and how quickly these are upgraded. and Josel (1981). It is clear that any savings vendor records 45(1) LRTS Book Vendor Records in the OCLC Database 19

may provide to libraries in terms of the keying process can eas- Handman, Gary. 1991. I lost it (or found it) at the movies: Public ily be consumed by the searching process when there is a high service implications of minimal-level cataloging of audiovisual rate of duplication. We have noted that further study needs to materials. Technical Services Quarterly 9, no. 2: 39–47. be done to determine the costs to libraries of documenting Horny, Karen L. 1991. What you see is what you get (so what do and reporting duplicates. Since OCLC’s deduplication pro- you get?): An administrator looks at minimal-level cataloging. Technical Services Quarterly 9, no. 2: 7–19. gram errs on the side of adding duplicates, OCLC should Intner, Sheila S. 1994. Taking another look at minimal level cata- offer reasonable compensation to libraries reporting dupli- loging. Technicalities 14, no. 1: 3–5, 11. cates based on costs determined by the recommended study. Johnson, Judith J., and Clair S. Josel. 1981. Quality control and the Recommendation 4: We strongly encourage OCLC to OCLC data base: A report on error reporting. Library consider increasing credits to libraries for upgrades, Resources & Technical Services 25, no. 1: 40–47. enhancements, and enrichments to records in the database. Marr, Russell. E-mail to the author on Oct. 26, 1998. Marr noted As more libraries move to systems that will allow them that it was not clear why the vendors lack access to these files. to download records from the catalogs of other libraries, and McCallum, Sally. E-mail to the author on June 5, 1997, as for- Internet searching of catalogs becomes easier, it is possible warded to the Program for Cooperative Cataloging list, that we will see a decrease in upgrading of any less-than- September 30, 1999, by Ana Cristan, Acting Team adequate records found in the bibliographic utilities. Giving Leader/BIBCO Coordinator, Library of Congress. O’Neill, Edward T., Sally A. Rogers, and W. Michael Oskins. 1993. significantly higher credits for upgrading and enriching the Characteristics of duplicate records in OCLC’s online union cat- database will give incentive for more member participation. alog. Library Resources & Technical Services 37, no. 1: 59–71. This credit needs to be concomitant with the cost of doing Patton, Glenn. 1991. OCLC’s long association with less-than-full the work. We determined that upgrading acquisition-level cataloging. Technical Services Quarterly 9, no. 2: 21–30. records was almost $25.00 per record. This falls within the Peet, Terry. 1999. Interviewed by L. Shedenhelm. SALALM range of $15.00 to $30.00 suggested by Steinhagen and Conference, Nashville, Tenn., June 3, 1999. Peet is the former Moynahan (1998). They go on to indicate that there is a head of the Hispanic Acquisitions Section at the Library of strong relationship between cooperation and economic Congress. reimbursement. To avoid undermining the cooperative spir- Preece, Barbara G., and Mary Anne Fox. 1992. Preliminary LC it behind internationally shared cataloging, more realistic records for monographs in OCLC. Information Technology and appropriately priced credits need to be given to the and Libraries 11, no. 1: 3–9. Romero, Lisa. 1994. Original cataloging errors: A comparison of libraries providing the intellectual content of the database. errors found in entry-level cataloging with errors found in Recommendation 5: We encourage vendors to take the OCLC and RLIN. Technical Services Quarterly 12, no. 2: opportunity to include value-added information in the 13–27. Compare to Cynthia C. Ryan (1978). records they produce. Ryan, Cynthia C. 1978. A study of errors found in non-MARC cat- John Riemer of the University of Georgia suggested aloging in a machine-assisted system. Journal of Library that vendors are in a unique position to give value-added Automation 11, no. 2: 123–32. Compare to Lisa Romero service at the point of record creation. If vendors have scan- (1994). ning equipment, they could readily add tables of contents to Sercan, Cecilia S. 1994. Where has all the copy gone? Latin their records with minimal increase of the labor costs. This American imprints in the RLIN database. Library Resources would provide additional information to prospective buyers & Technical Services 38, no. 1: 56–59. and act as an advertising technique, thus boosting sales. The Stamm, Andrea L. 1996. Minimal level cataloging: Past, present, and future. In Technical services management, 1965–1990: A added access would also give libraries more reason to look quarter century of change and a look to the future. Edited by favorably on vendor records. Linda C. Smith and Ruth C. Carter. New York: Haworth Press. Steinhagen, Elizabeth N., and Sharon A. Moynahan. 1998. Works Cited Catalogers must change! Surviving between the rock and the Beall, Jeffrey. 2000. The impact of vendor records on cataloging hard place. Cataloging & Classification Quarterly 26, no. 3: and access in academic libraries. Library Collections, 3–20. Acquisitions & Technical Services 24: 229–37. Wanninger, Patricia Dwyer. 1982. Is the OCLC database too large? Erbolato-Ramsey, Christiane, and Mark L. Grover. 1994. Spanish and A study of the effect of duplicate records in the OCLC sys- Portuguese online cataloging: Where do you start from scratch? tem. Library Resources & Technical Services 26, no. 4: Cataloging & Classification Quarterly 19, no. 1: 75–87. 353–61. Ferguson, Bobby. 1991. Upgrading minimal-level records for a Weitz, Jay. 2000a. E-mail to the author. June 9, 11:09 A.M. statewide database. Technical Services Quarterly 9, no. 2: 31–37. ———. 2000b. E-mail to the author. June 9, 11:58 A.M. Fox, Mary Anne, and Barbara G. Preece. 1991. Upgrading minimal level monographic records: A study and conclusions. Technical Services Quarterly 8, no. 4: 25–35. 20 LRTS 45(1)

The Effect of Interface Design on Item Selection in an Online Catalog David H. Thomas

The effect that content and layout of bibliographic displays had on the ability of end-users to process catalog information was tested using a 2 x 2 factorial experimental design. Participants were asked to perform two related tasks during the course of the experiment. In the first task, they were asked to select a set of items that they would examine further for a hypothetical paper they must write, using a simulated online catalog to make their assessments of relevance. In the second task, they were asked to examine 20 bibliographic records, decide whether they would choose to examine these items further on the shelf, and identify the data elements that they used to formulate their relevance decision. One group viewed bibliographic records on an interface similar to current online catalogs, one that used data labels and contained data elements commonly found. A second group viewed these records on an interface in which the labels had been removed, but the data elements were the same as those in the first. The third group viewed these records on a labeled display that included enhanced data elements on the brief record display. The final group viewed these records with the same brief record data elements as the third group, but with the labels removed, using ISBD and AACR2 punctuation standards. For the first task, participants using enhanced brief screen interfaces viewed more brief screens and fewer full screens than their counterparts. Screen durations for the second 10 screens were found to have dropped from those of the first 10 screens. Statistical analyses comparing demographic variables to the screen frequencies uncovered many significant differences. Participants using the enhanced-content interfaces made fewer selections from index and full screens, and more selections from brief screens. For the second task, participants who used enhanced-content interfaces were able to make some sort of relevance judgment more frequently than those who used standard-content interfaces.

he widespread introduction of online public access catalogs into libraries Tover the last fifteen years has had a major impact on the way that end users utilize libraries and the resources that they contain. The development of online catalogs has transformed the primary locating tool in libraries—the card catalog—from a tool with limited means of exploitation to one with a much greater potential to help users find needed information objects. Research has suggested, David H. Thomas (sunfish62@yahoo. however, that the online catalog brings with it a new set of problems for end com) is an independent Database users, as the added capabilities have also brought added complexities (e.g., Design Consultant. Larson 1991a). Migrating bibliographic data from a print to an online environ- Manuscript received February 22, 2000; accepted for publication June 23, ment has required reevaluation, redesign, and testing of many aspects of biblio- 2000. graphic information systems, from the methods of retrieval through the means of 45(1) LRTS The Effect of Interface Design on Item Selection 21

presenting this information. While a large number of helpfulness or obstructiveness of the library’s information researchers have examined functional aspects of the system- retrieval systems. However, the information system’s failure user interaction with online catalogs (e.g., Bates 1989; might occur because the system fails to communicate clear- Borgman 1986; and Harman 1992), few have investigated ly the content of the database. A failure of this type can the effect that screen content and layout might have on the occur because the user does not understand that the system effectiveness of the system-user interaction. The screen has successfully retrieved bibliographic records to meet the content and layout of online catalog interfaces have been user’s need. A system that presents the user with incomplete designed almost entirely based on expert opinion, with little or unclear data about the items it describes can mislead the use of empirical data on user preferences. user into thinking that the system does not contain items The purpose of this research was to contribute to the that meet the information need. For example, the data ele- knowledge about online catalogs in order to help improve the ment that caused a particular record to be retrieved in effectiveness of those catalogs. The primary objective was to response to an information seeker’s request may be embed- examine the communication process between the user and ded in a part of that record that the seeker does not initially the catalog by gathering basic empirical data about user per- see. This might occur, for example, when an individual formance on different types of online catalogs. This was done requests a specific work that is listed in a contents note and by testing whether the content and layout of bibliographic an added entry, neither of which the user sees on initial displays in an online catalog influenced the effectiveness of screens. In other words, how the system presents its findings the interface between online catalogs and end users. to the user can influence the success of the communication process in an information retrieval exchange. This problem underscores the fact that an information The Information Seeking Process storage and retrieval system is, in fact, an information storage, retrieval, and presentation system. It is not enough for In the information seeking process, an individual lacking the system merely to store and retrieve information; to be knowledge to solve some need or problem seeks to remedy useful, this information must be presented to the user in a that lack by obtaining knowledge from some information manner that the individual can interpret meaningfully. resource (Buckland 1991). Such resources can range from The adoption of cataloging standards, such as the Anglo other people (e.g., authorities) to museums or libraries. American Cataloguing Rules, 2d. ed. (AACR2), represents a When the resource that the user consults is a library, the recognition of the user’s need for standardized bibliograph- individual may employ the library’s locating devices to iden- ic descriptions. Without such standards, the user would have tify potential informing instruments. The mechanisms in a to interpret each library’s catalog separately, which can hin- library by which an individual can obtain needed informa- der the communication process. Similarly, the International tion include consulting a reference librarian, reference Standard Bibliographic Description (ISBD) represents an sources, or the library’s collections. In the latter two cases, effort to establish guidelines for the arrangement of biblio- the individual often must use the library’s bibliographic cat- graphic data in a unit description, in part to make it possible alog to locate the specific informing entities (e.g., almanacs, for the user to interpret the contents of the description more encyclopedias, journals, monographs, etc.) that are needed. easily. However, because the library serves the needs of many With the advent of online catalogs, awareness of the different users, its catalog contains data designed to address user’s need for standardized bibliographic description seem- many potential problems or needs, and the process of locat- ingly waned. This might have been due in part to the fact ing useful informing entities for an individual’s particular that the complexity of introducing computer technology in need can be difficult and frustrating. Individuals might seek bibliographic applications distracted system developers, and information about particular attributes of a library’s holdings, it might have been due in part to the desire of developers to as when they look for items created on or after a certain date; experiment with the power that the new technology offered. they might require a particular bibliographic entity in the Moreover, the ability to manipulate the content of the online collection; or they might need information of a more gener- catalog more freely has given the library profession the abil- al topical nature. Moreover, individuals must try to match ity to question the bibliographic display content guidelines their conceptualizations of the problems with the terminolo- previously established. The wide variation in current online gy used by the information system to represent those prob- catalog screen designs supports this view. lems. These factors make a complicated process of bringing One negative result of this experimentation has been the information seeker together with the information sought. the loss of familiarity for the user, as each library’s catalog This search process can be made easier or more difficult can have different features and displays. The user is once depending on many conditions, ranging from the ability of again placed in the situation of having to learn how each individuals to articulate their problems to the degree of library’s catalog functions and how to interpret the output of 22 Thomas LRTS 45(1)

each of these systems. Different displays may contain the in the exchange between the online catalog and the library same kinds of data, while seemingly similar displays may user. Indeed, the need for ongoing research into this inter- have different information. Such confusion makes using action has been recognized for some time. For example, online catalogs more complicated for the individual. Cochrane and Markey (1983, 340) list “analyzing user An examination of the library literature reveals that the requirements and behavior” as the first of four priorities in content and layout of few—if any—online catalog displays need of systematic study with online catalogs. In response to have been rigorously tested to determine whether they meet these perceived needs, a large body of data has been gath- users’ needs, and that most catalogs have been designed ered regarding online catalog user behavior, but few authors without such input or testing. In human factors research have examined user preferences for online catalog displays. into computer screen design, researchers have established Human factors research is more varied, as the number and broad guidelines for efficient and clear screen design— type of applications. It ranges from studies in which authors broad guidelines that library systems designers have used to examine the effect that particular design attributes, such as inform online catalog designs. However, these researchers, color or highlighting, have on interface usability, to broad- too, have largely failed to base screen display design princi- based handbooks on general interface design. This literature ples on empirically gathered data. This might be because includes numerous empirical studies that might inform the wide variation in applications makes such detailed study online catalog design. meaningless, inasmuch as the layout of data onscreen is so The efficiency of the information exchange between the dependent on the purpose and content of the data. Library user and the information system is a major concern in this catalogs are bounded, however, by function and content. By study. The central question is whether participants perform and large, they serve the same purposes (identification and better when using screens with particular data on them in location of information objects), use the same data structure particular layouts. Consequently, this review begins with a (the MARC format), and use the same data content (as pre- description of research about library catalog performance. scribed by cataloging rules such as AACR2). Also examined is literature about human factors research In addition, bibliographic data differ from typical com- and interface design, and cognitive aspects of users’ utiliza- puter data structures. Examination of manuals of screen tion of an interface. Researchers interested in a more design suggests that most applications use data of repetitive detailed review can see Thomas (1997). content and fixed length. In contrast, bibliographic data are Authors in each of these areas consider the human- highly variable in content and length, thus making the whole- computer interaction from different perspectives. Those sale application of these general guidelines difficult. For who conduct system-oriented research examine this interac- example, labeling the data elements of even a simple MARC tion in terms of information system features, and attempt to record can be challenging, because different types of data determine whether the system provides the user with appro- can reside in the same position of a display—as in the priate functions and usability. This is relevant because instance of the main entry, which can be a personal name, a screen design is integral to system design, and tests of sys- corporate body, a conference name, or a uniform title. tem performance can include tests of screen layout. Some Furthermore, the existence of pre-established formatting researchers in human factors attempt to view more specifi- standards for bibliographic data frequently render conven- cally the effects of the presentation of system content and tional design formulae inapplicable. ISBD, academic citation function on the human-computer interaction. Authors who forms, and the catalog card—all of which existed prior to the study cognitive aspects of information system components introduction of online catalogs—are each familiar, alternative seek to test the human-computer interaction in terms of means of presentation of bibliographic data. These two fac- how the brain processes information. Most of the tors—the nature of bibliographic data and alternative means researchers in each of these areas consider only the ability of of formatting bibliographic data—suggest that standard the system to communicate to the user the structure and screen design criteria should be examined and tested. functions of the information system, however, and do not consider the effect of content presentation on the efficiency of the human-computer interaction. Literature Review System-Oriented Research The literature of human-computer interaction is large and diverse, and researchers in both the field of human factors With the introduction of online catalogs to the library, the and that of library and information science have studied field of catalog use research has flourished. Computer-based many aspects of interface design. Researchers in the library systems have made it possible to gather data on users’ search- and information science field have been concerned with the es that could not be gathered in the manual catalog environ- human-computer interaction specifically as it is manifested ment. This capability has given librarians and researchers an 45(1) LRTS The Effect of Interface Design on Item Selection 23

unprecedented opportunity to examine what users do with face. Nonetheless, Harman concentrates on improving the catalogs. Investigations of catalog use are relevant here functional aspects of the catalog—on improving search capa- because in this study the communication process between bilities—rather than on improving displays. While these sys- the user and the system is examined in an attempt to analyze tems do offer a range of different display types, there appears sources of communications breakdowns. With regard to this to have been little testing of the screens on users, and thus lit- research, however, most researchers have sought ways to tle effort to determine whether they present the user with improve the search engine or the database mechanics, rather information in a more comprehensible manner. than questioning the content of specific displays. Similarly, the majority of researchers who have gath- Two aspects of catalog use that have received a great ered empirical data on user actions at online catalogs have deal of attention from researchers are subject searching and attempted only to document the functional aspects of a sys- indexing, in large part because of their problematic natures tem that are used, and not how users evaluated the data pre- (Larson 1991a; Markey 1985). Subject searching has been sented to them or whether they understood the meaning of identified by users as the most difficult aspect of catalog use. these data. Both Borgman (1988) and Seymour (1991) pro- Despite this, subject searching is one of the most frequent- vide reviews of this literature. Seymour arranges them by ly used methods of searching (Bates 1986). Authors of many the method of data collection, with categories for surveys, such studies consider the mode of presentation of this infor- experiments, transaction log analyses, interviews, and direct mation to be substantially deficient. Larson (1991a), in dis- observation. In addition to the method of data gathering, cussing the problems of zero retrievals and information Borgman also attempts to identify and categorize the vari- overload, envisions a future interface with user-definable ables studied. In this list, Borgman includes as a final vari- screen formats and functions. Similarly, Bates (1989) pro- able for study the “effect of different interface methods poses an interface that uses a “Super-Thesaurus” to assist (e.g., command, menu, form fill-in) on behavior,” but then users in entering the catalog. Neither Bates nor Larson give notes that this area “has not been studied empirically for details of how such innovative systems might be presented information systems” (145). The need for research into to the end user, although Larson does consider, in his dis- mode of presentation is not raised. cussion of methods for remedying subject search problems, Examples of use studies of online catalogs include: the catalog interface to be one important factor in the over- Larson (1991b); Millsap and Ferl (1993); Hancock-Beaulieu, all success of the information transfer process. Robertson, and Neilson (1991); and Siegfried, Bates, and Similarly, Borgman (1986) examines the difficulties that Wilde (1993). The authors of these studies examine aspects online retrieval systems pose to end users. Borgman identi- of user online behavior, and typically offer suggestions on fies numerous problem areas that have been studied, group- how these aspects might be changed or assisted in some way. ing these into problems with mechanical aspects of retrieval Larson (1991b) used nearly complete transaction logs of a systems and problems with conceptual aspects of the process major academic library system for a six-year period to exam- of retrieval. She reviews the research concerned with the ine changes in search patterns. Using logs for the University effect that individual differences play in this process, noting of California’s MELVYL online catalog, Larson found that that most researchers have found that the search process dif- subject searching over the period declined by about 2% per fered depending on cognitive style, but that search results year as a percentage of searches performed. He suggested did not. Other researchers have found that overall experi- that this decline was the result of an increase in title word ence with retrieval systems affected search results. Borgman searching, as users moved away from the complexities and points out, however, that little research has been conducted failures of subject searching. Larson recommended that into the effect that interface design has on the success of online catalogs offer users assistance in mechanical aspects of searches. The only authors whom Borgman cites who are searching, such as partial matching and stemming of terms, concerned particularly with online catalog interfaces, Martin but he did not question whether the mode of presenting this (1974) and Hildreth (1982), list online catalog features, but information might be improved. do not evaluate how these features might be valued, inter- Millsap and Ferl (1993) examined the transaction logs preted, or used by the end user. of more than 1300 remote uses of MELVYL and found that Harman (1992) places a different emphasis on this issue, a large number of searches were basic and retrieved one or and questions whether changes in the user interface can make more citations. However, a substantial minority of users had a complex Boolean retrieval system truly easier to use. At their difficulty using the system, retrieving either no entries, or most basic level, the author argues, these systems require the unmanageably large numbers of them. Again, the authors end user to understand the theory behind Boolean logic, a proposed adding online catalog features that would assist requirement that she implies is too demanding. The author the user with mechanical aspects of searching. then describes four prototype search systems based on statis- Siegfried, Bates, and Wilde (1993) explored the search tical retrieval techniques, each with a custom-tailored inter- behavior of humanities scholars through transaction log and 24 Thomas LRTS 45(1)

protocol analysis. In their conclusion, they stated (288) that make them clearer significantly reduced decision-making their “results suggest that the design of database search serv- time for users. From this, Tullis (1983) developed a means ices . . . is still far from optimal for meeting the needs of of measuring the complexity of screens based on groups humanities scholars.” As before, the authors’ main concerns defined by character proximity. Marcus (1982) discussed the were with finding means of improving the mechanical aspects role that graphic design principles played in the design of an of searching, and not the content of the user interface itself. interface for a geographic information management system. Thus, researchers in library and information science Kruk and Muter (1984) tested several aspects of reading text have focused largely on improving and documenting on video screens. The researchers found that reading text on mechanical aspects of these systems. Most researchers would video screens was slower than from printed matter, and that probably agree that online interfaces continue to need the vertical spacing of text on screen affected reading speed, improvement; however, few have suggested that screen con- but that screen contrast and distance from the screen had no tent or layout could play a role in this improvement. effect on reading speed. Trollip and Sales (1986) tested the effect that fill-justi- Human Factors Research fied text (i.e., text that has even margins on both sides) had on reading speed and comprehension. They found that read- Researchers into the human-computer interaction at online ing speed was significantly slower for fill-justified text than catalogs are acutely aware of the limitations of current-gen- for left-justified text, although comprehension did not differ. eration online catalogs, but their focus in finding solutions to These studies, which were concerned with issues related to these problems is on making the functions of the online cat- the impact that the layout of data has on interface effective- alog easier and clearer for end users. In the developmental ness, might guide library systems designers in designing their stage of a new technology such as online catalogs, this focus displays, assisting with some of the many design decisions is understandable and warranted. However, as online cata- that must go into a typical online catalog display. logs have become easier to use on a functional level, it has Tullis (1988) describes research conducted to identify become increasingly important to test the ease of use of cat- and test the correlation between subjective and objective alogs on an informational level—i.e., to test the means of measures of screen complexity. In a preliminary experiment, conveying database content to the user. Library and infor- Tullis isolated the correlation between these two measures mation science professionals might look to the field of by presenting participants with various screen displays and human factors research to find a body of research that can having them give a subjective rating of the screens. help inform many aspects of the design of online catalogs. Participants were timed during the task, and these times Human factors research is a broadly defined field that were used to correlate the measures of screen complexity examines many aspects of the interaction that humans have with time to complete the task. Tullis found that both time with many different technologies. One facet of human fac- to complete the task and subjective rating were positively tors research, however—that of computer interface correlated to screen complexity. In a follow-up experiment, design—has attempted to examine precisely some of the Tullis used the results of the first experiment to predict the concerns that library and information science professionals outcome of a similarly structured experiment. Tullis found a need to consider. Human factors research into computer high positive correlation between time and complexity, interface design will be examined here, with particular while the subjective rating/complexity correlation was not attention paid to the issue of screen layout and the effects significant. The researcher then applied the evaluation cri- that screen layout might have on system usability. teria to earlier experiments of screen display, and found high Several authors have written textbooks devoted to the correlation between the predicted values found by his crite- design of user interfaces (e.g., Shneiderman 1992; Galitz ria to the actual findings of those earlier studies, which lends 1989; Brown 1988; Dumas 1988; Crawford 1987; Norman support to the use of these criteria. 1990; Rubin 1994). Although the authors of these volumes Tullis (1988) offers several design criteria that could be support most of their design principles with empirical used to test alternative bibliographic displays. Tullis found research results, this support weakens when each discusses that the number and size of the groups of elements onscreen layout and content choices. Instead, the authors tend to offer had the greatest impact on search time, while the density of common-sense guidelines when discussing screen layout. characters onscreen and the layout complexity (as defined The authors of these general texts rely largely upon common by vertical alignment of screen elements) had the greatest sense because there has been little empirical data gathered impact on subjective rating. These criteria might well be on the effect of a screen’s layout on system usability. applied to bibliographic displays, although the complex Examples of empirical research in screen design nature of bibliographic data and their potential for great include the work of Tullis (1981; 1983). In the first of these length would need to be accommodated. Specifically, Tullis studies, Tullis (1981) found that reformatting screens to identifies methods for grouping data elements on-screen, 45(1) LRTS The Effect of Interface Design on Item Selection 25

assuming that each element will occupy the same amount of online catalog screen displays based on guidelines from space in every case—something that can’t be assumed with human factors literature, again derived from common sense. bibliographic data. In online catalogs, data layout decisions Shires and Olszak (1992) reviewed basic interface design must either be made once on a best guess basis, or for each issues for online catalogs and focused almost exclusively on record as it is prepared for display. Online catalogs to date screen design for conveying command and search concepts have used the former method. more clearly. Yee (1991) surveyed research on user inter- It is perhaps understandable that authors of human fac- faces in order to identify research methods and findings tors research—with the exception of Tullis—have given applicable to the design of effective user interfaces to online such scant attention to testing particular aspects of screen catalogs. On the issue of single record display, Yee points out layout. Screen utility is closely related to the function of the that many have asserted that labeled displays aid in user application, the data that the system must convey, and the comprehension, but only one group of researchers has in purpose to which the user will put the information. Given fact studied this phenomenon. the variation in computer applications, such studies would In a handbook for library interface design, Crawford likely be limited in their generalizability, and of only minor (1987) relies almost exclusively on his own judgments of interest to nonusers of the particular application. Indeed, interface design upon which to base design decisions. In this where there are areas of general concern, such as optimal volume and in Crawford, Stovel, and Bales (1986), Crawford menu design or the relative benefits of menu and command asserts the superiority of labeled displays without offering interfaces, there exists a body of research. For the purposes evidence to support this choice. He does examine this issue of this research, however, an examination of the human fac- in some detail, noting, for example, the fact that choosing a tors literature offers little guidance, because the authors of citation format is complicated by the lack of standardized this literature mainly focus on conceptual and functional citation formats. In addition, he attempts to address some of aspects of interfaces, whereas this study has been focused on the problems that bibliographic data commonly present in those aspects of the interface that facilitate communication designing displays, especially the variability and complexity of information between system and user after the user has of the data, and its potential size. While the design princi- already navigated the conceptual barriers. ples that Crawford presents are sound, there is no attempt Other human factors researchers have conducted to test these principles empirically. research into the mechanical aspects of the computer inter- Crawford, Stovel, and Bales (1986, 2–3) identified five face, examining the effects that various interface features questions of importance to the design of bibliographic have on user performance. However, the authors of these displays: studies, like those in the library domain, have focused primarily on the mechanical or functional aspects of the interface, 1. Does the display provide an appropriate amount of and not on the effect that layout might have on information information? transfer. Even so, such studies underscore the fact that display 2. Will patrons understand the information as it is dis- factors can affect user performance, even if the researchers played? do not generally address questions of screen layout. 3. Is the display readable and attractive? In the more specialized area of library information sys- 4. Will patrons be able to find information rapidly and to tems, the literature is sparse and focused on content rather find all the information needed? than presentation. Shaw (1991) reviewed the recent litera- 5. Will patrons be able to view the information on a sin- ture in this area, but did not discuss any studies that test the gle screen? effect of layout on performance. It is interesting to note that several authors cited by Shaw noted the need for empirical The first four questions deal with important display evaluation of user interfaces. In a study of card catalog use, issues regarding the human-computer interface. The authors Palmer (1972) found that most users concentrated attention state that most work on interface design has been focused on author, title and call number data; subject headings were primarily on the first, and to a lesser degree the second, used by just under half of the users. From this, Palmer sug- third, and fourth questions. The authors then examine the gested that a five-item catalog using author, title, call num- more practical fifth question. They do not attempt to address ber, subject headings, and date of publication would meet these first questions further; instead, they use aesthetic judg- the needs of a substantial majority of the users’ requests. ments to guide discussion of screen design. They then pres- Seal (1983) reported that 90% of studied users were sat- ent a number of display options, and assert the relative value isfied with a short entry catalog. Hufford (1991) found that of these screens without offering any empirical data to sup- the reference librarians he studied used only particular data port their claims. At no time do the researchers offer empir- elements in most situations, confirming the results of earli- ical evidence to support their choices of: fields to display, er studies. Matthews (1985) provided a set of guidelines for labels to use, or orders of information to present. Thus, 26 Thomas LRTS 45(1)

although the authors discuss the relative merits of differing uses three basic assumptions about information seeking, two display types, they do so from a perspective limited in its of which have bearing on the current research: first, that end basis to personal opinion and not supported by the results of users aim to solve a problem, not use the information any wholly objective assessment. retrieval system; and, second, that end users seek to reduce In a different vein, Smiraglia (1992) provides a helpful the cognitive load of using an information retrieval system. outline of the basic goals of the catalog, which he labels This second observation is supported by Akeroyd (1990) as identification, collocation, and evaluation. This outline well as Tullis (1988). Marchionini, Akeroyd, and Tullis all might be used to help inform interface design. The first two seem to suggest that the online catalog interface should functions, identification and collocation, depend upon pre- facilitate the search process in part by presenting biblio- senting traditional bibliographic data elements. As Leazer graphic data to the user in the most effective manner. (1993) notes, Smiraglia’s identifying function closely paral- Branching off the basic research in interface design is a lels those of Cutter (1904) and the Paris Principles (1963)— broad area of research into prototype interfaces and system that is, identifying particular items in a library collection. innovations. The authors of these studies typically take as Two parts of Cutter’s first function of the catalog—those of their starting point an acknowledged failure of current inter- identifying whether a library has any given item by an author faces, and devise some new means of providing the user or with a particular title—clearly have as their primary focus with system features. As with system-oriented studies, how- the bibliographic item. The third part of this function— ever, many authors (e.g., Frei and Jauslin 1983; Davis and identifying whether a library has a book on a particular sub- Shaw 1989; Noerr and Bivins Noerr 1985; and Belkin, ject—does not focus as directly on the item. When users Marchetti, and Cool 1993) focus on simplifying search func- seek material on a given subject, they do not usually have tions, and not on clarifying screen content. particular items in mind. Because of this, this part of the Authors in one area of specialization do concern them- function should perhaps be considered separately. For the selves to a great extent with the best means of presenting first two parts, traditional bibliographic data elements— information to the user. These researchers, who are all author, title, and publication information—might sufficient- studying the problems associated with the visualization of ly serve in most cases, for specific item searches. information, have begun to examine whether modes of com- The collocation function seeks to bring together items munication other than textual communication might facili- with a particular attribute, e.g., author or subject. As with tate information retrieval. Olsen et al. (1993) presented one the identifying function, traditional bibliographic data ele- such system, VIBE, where users perform an information ments serve this function in most cases. The most notable retrieval task using a two-dimensional plane with “Points of failure here is the lack of topical data elements included Interest” that can be manipulated to find relevant materials. among most traditional bibliographic data elements. Beheshti (1992) outlined another such system, where Moreover, this function is more complex, in that the nature MARC records are rendered as virtual books on shelves that of the use is less precise. Users who seek to know what a are displayed to the user on screen. The researcher envi- library has on a particular topic or by a particular author sioned this system enhancing users’ abilities to browse have a less precise knowledge of the items they seek than library collections. In recent years, this area has received those who seek specific items. This places different substantial attention, and offers promise for alternative demands on the information system, and requires that the means of informing users of database content. system provide enough information to the user that an eval- In summary, then, researchers in the human factors uation can be made. area, both within and outside the library field, have explored The third function, the evaluative function, is problem- many features of interface design. The problem remains atic in that it is highly situational. It is the point at which the that, in addressing the issue of screen layout, a great deal of relevance of retrieved items is determined. Relevance is one this literature is not based on empirical research, but rather primary means of measuring the effectiveness of an infor- on aesthetic judgments or common sense. There are notable mation retrieval system. However, the data necessary for exceptions. For example, Tullis’s work does present such deciding upon the worth of a given bibliographic item is empirical evidence. Unfortunately, this research has not dependent on the user’s particular information need, a need been replicated using bibliographic data. that is likely to be poorly defined and subject to constant change. Indeed, the detailed study of the particular data ele- Cognitive Research ments that any user might need at any given time could form the nucleus of a large body of research, which is beyond the Research into cognitive issues in information systems could scope of this research. lead to the development of predictive models of specific dis- Marchionini (1992) outlines the information retrieval play need. In one area of this field, researchers attempt to process and features of systems to support that process. He create models of the cognitive makeup of users, so that sys- 45(1) LRTS The Effect of Interface Design on Item Selection 27

tems might be designed to predict the particular needs of argued that establishing a system that accounted for cogni- users in given situations. Individual user profiles can then be tive style would be advisable, but that such a system should created by building on these base models, thus enabling the leave the choice of mode up to the user. However, asking all system to anticipate an individual’s preferences, and to mod- users to choose the search mode for a session seems cum- ify the model iteratively. One aspect that these systems bersome and requires users to be able to identify their own could manipulate in response to these models is the screen preferred style. The reader will recall, moreover, that design and layout. For example, an expert system designed Marchionini (1992) suggested that users do not care about to use cognitive models could provide greater or lesser how a system works, just that it works. screen complexity depending on a number of user variables. Allen (1994) used an experimental method to perform Morehead and Rouse (1982) and Brajnik, Guida, and two tests of the relationship between a cognitive ability and Tasso (1987) explored the use of such user models in infor- the ability to perform a retrieval task. In the first test, Allen mation retrieval systems. Aykin and Aykin (1991) reviewed tested the relationship between logical reasoning, as demon- the literature of individual differences and their effect on strated on a standardized test, and ability to identify relevant human-computer interaction. The authors pointed out in citations from two online catalog displays, one sorted in their conclusion, however, that “there are only a few stud- reverse chronological order, the other in relevance order. ies considering a limited number of user characteristics” The researcher found a significant interaction between log- (377). Although the authors of much of this research again ical reasoning ability and the type of display, and particular- focus on user models to predict system functional needs, ly that users with lower logical reasoning ability benefited research into screen design could potentially be affected by most from the relevance order display. The second experi- such considerations. ment tested the relationship between perceptual speed and Several authors have examined the impact that cogni- the ability to examine subject headings arranged either tive differences between individuals can have in the infor- alphabetically or hierarchically. In this experiment, the mation exchange process. Hensley (1991) offered researcher found no significant results. However, users of suggestions on how reference librarians can use learning the hierarchical display examined substantially fewer lines to style theory to facilitate the reference process. Prorak, identify relevant headings, which might indicate that the Gottschalk, and Pollastro (1994) tested the effect that teach- method of presentation affected participants’ perception of ing method has on teaching bibliographic instruction; the the system. Both experiments, then, offered evidence that researchers found no significant preferences. Allen and the interface can affect usability of the system. Allen (1993) tested the relative cognitive abilities of librari- Kerr (1990) tested the impact that different screen ans and students, and found significant differences. They enhancements have on assisting users in navigating through found that librarians had higher logical and verbal compre- electronic information systems. The author found no signif- hension capabilities, while students had higher perceptual icant difference in usability between systems with no screen speed. The authors considered that these differences might enhancements and screens with enhancements of color, suggest that individuals use information retrieval systems icons, textual headers, or a combination of the three. Coll, differently as a result of cognitive differences. Again, these Coll, and Nandavar (1993) presented research designed to researchers did not examine the effects that these cognitive test whether good screen design depends on layout or con- differences had on screen preferences. ceptual considerations; the researchers found that concep- Ford and Ford (1993) tested whether different cogni- tual structure plays a greater role. Participants were asked to tive strategies affect success in addressing an information choose items from menus that were arranged alphabetically, need. The researchers asked students to learn about a par- categorically or randomly, and on which the function key ticular indexing system using a knowledge base that includ- assignments changed from screen to screen. They found ed (unknown to the participants) two human experts. An that task times for the two organized menus, while similar to examination of the search sessions showed that participants one another, were significantly shorter than the unordered asked different types of questions of the knowledge base, menu. This study reinforced the idea that how data are dis- and that the question types did affect problem-solving suc- played can affect system performance. cess. The researchers found that two approaches—one asso- Overall, the authors of the literature reviewed in this ciated with female participants and focused on detailed section focus primarily on testing aspects of the functional information, the other with males and focused on broader human-computer interface. Library researchers have exten- analytic information—led to success, while a third, which sively examined how users search and the errors that they focused on middle-level concepts, did not work. In their have made, through transaction log analyses, user surveys, conclusion, they linked the two successful approaches with interviews, and other methods. A great deal of attention has specific cognitive styles identified by earlier researchers: been paid to the complexities of the search and retrieval operation learners and comprehension learners. They then process, with many studies comparing the performance of 28 Thomas LRTS 45(1)

different search methods. Similarly, researchers in human Mandel 1991; Mandel 1985). Such reassessments have been factors have tested the relative usability of command-driven hindered by the lack of data concerning such user prefer- and menu-driven interfaces, and the best means of organiz- ences and needs. This study, while not aimed primarily at ing items in a menu. Some research into overall screen dis- determining the importance of particular data elements in play has been conducted, and some measures of screen bibliographic databases, might nonetheless provide a base- complexity have been devised, but these tests and guidelines line of preferences upon which others might build. have not been tested with bibliographic information. Identifying individuals’ preferences on display screens might provide a new perspective on which data elements they consider important to their problem-solving process. Importance of the Study It has been suggested that the need for evaluating user preferences in online catalog displays will soon be obviated Because end user access to remote databases of all types is in the client-server environment, where an individual will be rapidly increasing, library systems designers must assume able to configure the database view in any preferred format greater ability to be responsive to the end user’s needs. In a (e.g., Larson 1991a; Buckland, Norgard, and Plaunt 1993). remote access situation, the user must rely entirely on the While end user configuration might alleviate some problems system designer’s ability to present the system to the user of choice of sorting and display, it seems unlikely that it sensibly; professional librarians cannot interpret displays. A would help the casual user, who must use the online catalog library system that presents end users with data that are “out of the box.” Moreover, many users will not make the more directly related to the needs of the user’s problem- effort to customize their displays. For such users, default solving process and are easier to interpret will require less displays must exist to present online catalog results. explanation to the end user. As electronic resources become In the near term, it will continue to be necessary to pro- more central to library operations, libraries’ overall effec- vide all users with fixed display types, and common sense tiveness will be judged in large measure by the performance suggests that these displays should provide information to of their information systems. the end user in the most sensible context. Externally creat- To design systems that better serve the end user, it is ed limitations force users to adapt their needs to the system, essential that library and information science professionals and the system does little to mitigate the user’s interpretive understand more precisely what users need to see onscreen burden. Every such burden compounds the user’s negative and how they want to see that information. While it is associations with the online catalog, while reducing this bur- important to design more sensibly organized interfaces on a den increases both the chances of a user’s success and the functional level, the question of display nonetheless remains user’s positive associations with that system. An aim of the critical—especially with data as complex as those found in a research reported here is to establish a basis for the design bibliographic database. An interface that offers information of more effective online catalog displays. that is perceived as more pertinent will be less obtrusive, and will thus enable the user to focus on the primary task (Benbasat and Todd 1993). Method It has been claimed (cf. e.g., Crawford 1986) that users of bibliographic systems are unaware of the content of particular A 2 x 2 factorial experimental design was used. The partici- screens, and that they cannot distinguish the different ele- pants, first-year students in a four-year undergraduate pro- ments of a bibliographic record. While it might be true that gram at a large university, were randomly assigned to one of typical end users do not consciously make these distinctions, four groups and then pretested for basic demographic infor- they nevertheless might be aware of them on some level. It mation. They were asked to perform two related tasks dur- behooves librarians to make their systems as clearly organized ing the course of the experiment. In the first task, they were as possible, both from functional and communications per- asked to select a set of items that they would examine fur- spectives. This research was conducted in order to provide ther for a hypothetical paper they must write. In the second baseline data that could make it possible to determine how the task, they were asked to examine 20 bibliographic records, content and layout of interfaces might be changed to make decide whether they would choose to examine these items bibliographic displays easier for the end user to interpret. further on the shelf, and identify the data elements that they Gaining an insight into the display needs of library cat- used to formulate their relevance decision. In both tasks, the alog users should help librarians make informed decisions topic of the task was the same: big band music and the music about changes to catalog content and structure. A number of of Duke Ellington. The instructions for forming relevance library professionals have called for a radical reassessment of judgments in each task were also the same: participants cataloging practices, so that both the MARC format and the were asked to use only the information provided on screen cataloging process might be simplified (e.g., Gregor and to make relevance decisions. 45(1) LRTS The Effect of Interface Design on Item Selection 29

Each group was presented with a simulated online catalog that listed a set of bibliographic records sorted in reverse chronological order, from which they had to make their choices. The listing for each group differed in the data elements and screen layout used, but contained the same bibliographic information. That is, they all viewed the same bibliographic records, laid out in different on-screen config- urations. One group viewed bibliographic records on an interface similar to current online catalogs, one that used data labels and contained brief display data elements from MARC tags 1xx–3xx. A second group viewed these records on an interface in which the labels had been removed, but the brief display data elements were the same as those in the first. The third group used an interface that employed tags, but in which the brief display data elements were changed so that they included more subject-oriented fields (omitting main entry and Figure 1. Single Line Listing, Used on All Interfaces physical description fields) while the full display was based on ISBD data elements. The final group viewed these records with the same brief record data elements as the third group, but with the labels removed, using ISBD and AACR2 punctuation standards. Using both tagged and untagged interfaces for each of the two content types helped control for differences in performance caused by particular layouts. Samples of these screens are provided in figures 1 through 9. Retrieval speed and comprehension were tested in the first task by gathering transaction data for each participant. The system logged each participant’s actions: the screens that were viewed and the time spent viewing them. Quantitative data were gathered on: the number of screens viewed; the number of citations examined at each level of specificity; the time to complete the task; and the time spent viewing particular screens. Bibliographic records for the experiment were gathered by searching the Research Libraries Information Network (RLIN), the OCLC Online Union Catalog (OLUC), the Figure 2. Standard Content, Labeled Brief Display Library of Congress catalog, and the University of Pittsburgh catalog. General searches were made for the key- words “Duke Ellington” and “Big Band Music”—both terms experimental interfaces. Because the number of data ele- taken directly from the problem statement. Duplicate ments included in the MARC formats numbers in the hun- records were removed from the master list, under the dreds, an attempt to test the relative value of all different assumption that a single library system would have one cat- combinations of these elements would have been impossi- alog entry for each of these items. 103 bibliographic records ble. Previous research had shown that most users tend to were used in the experiment. To populate the second task, a employ only a limited portion of the bibliographic record random set of 20 records was separated out, leaving the first (Palmer 1972; Seal 1983; Hufford 1991), and that briefer task with a list of 83 records. For the second task, the 20 catalogs containing only a few of the data elements current- records were randomly ordered and randomly assigned to ly found in bibliographic databases would serve a substantial display in full or brief layout. majority of end users’ needs. Based on these earlier studies, brief display screens for this experiment were designed that Screen Design Issues included data elements from a limited set. One method for choosing alternative data element sets The choice of screen elements for use in the different dis- for display to users might be to examine the data elements plays posed an important problem in the design of the in a bibliographic record in relation to some of the accepted 30 Thomas LRTS 45(1)

Figure 3. Standard Content, Labeled Full Display Figure 4. Enhanced Content, Unlabeled Brief Display

Figure 5. Enhanced Content, Unlabeled Full Display Figure 6. Standard Content, Unlabeled Brief Display functions of the catalog—most notably Cutter’s definitions a particular group—it might be possible to test whether the (Cutter 1904) and those found in the Paris Principles content and layout of a screen has any effect on participants’ (International Conference 1963). Based on such a function- performance in that given situation. al analysis, a set of data elements can be chosen for inclusion The evaluative function was chosen because it would in particular displays. These displays would then ostensibly force the participants to rely more heavily on the informa- support the stated functions of the catalog. tion provided. Users who seek a particular known biblio- This functional approach was used, and provided a basic graphic item in a collection might not seek to view the same framework within which data elements were selected for data elements as those users with a more generalized infor- inclusion on the bibliographic screens. In this research, mation need. Similarly, users who seek quick answers to attention was focused on the evaluative function of the solve an information need will potentially need to view dif- online catalog by inducing the participants to use a topically ferent data elements (in this case, perhaps, the extent or oriented search strategy. By controlling the content of the availability of the bibliographic entity). For the purposes of user’s information need—as in assigning an artificial task to this research, participants were given a topic of which they 45(1) LRTS The Effect of Interface Design on Item Selection 31

Figure 7. Standard Content, Unlabeled Full Display Figure 8. Enhanced Content, Labeled Brief Display

whereas those who seek to meet a more visceral need must evaluate the bibliographic data provided to them more closely. As a result, such users are more likely to rely heavily on the data presented by the system to determine the worth of those items for solving the information need. Such use places greater demands on the information system, which is the reason why this type of situation was established for this research. In order to conduct this research, content designations for three screen types were designed: one that listed each item on a single line; a second that listed an individual item using a brief citation; and a third that listed the full contents of an individual item. These screens were designed based on the results of prior research (e.g., Palmer 1972; Seal 1983; Hufford 1991), as well as on library standards (e.g., ISBD and AACR2). For the single line display, brief title (MARC 245), author (MARC 1XX), and date information (MARC Figure 9. Enhanced Content, Labeled Full Display 008) were used; for the treatment group brief display, title (245), edition (250), publication information (260), subjects (6XX), and call number (090) were used; the treatment presumably had little foreknowledge, and were asked to group full display included all MARC fields. compile a list of most useful sources based on the information provided in the information system. Administration of the Test In known-item searches, users seek to verify or rein- force certain facts that are already known. In contrast, sub- Participants volunteered for one of eight sessions over two ject-oriented searches are more tentative and probing in days. Due to hardware limitations, 8 groups of 16 partici- nature. The user seeks to discover whether the library owns pants were scheduled for the experiment. Within each ses- bibliographic entities—or more generally whether there sion, all four interfaces were tested, with each interface exist any bibliographic entities—that might meet an infor- being assigned to an equal number of terminals, which were mation need. This more diffuse need places different grouped together. Every participant thus had an equal requirements on the system. Users seeking known biblio- chance of using any of the interfaces. They were grouped graphic items can rely on their ability to differentiate pat- together to minimize the chance that neighbors would terns of data and match on incomplete user knowledge, notice differences in the interfaces. 32 Thomas LRTS 45(1)

Participants were instructed to seat themselves at any age and a spreadsheet program. Tests used for the analysis terminal. Each interface used the same opening screen, which of the data are listed here, along with the identifying label eliminated the possibility that participants might choose one used to state statistical results: student’s t-test (t); Chi- interface over another based on some screen difference. squared (χ2); One-way Analysis of Variance—ANOVA (f). The use of several different administrations of the test The statistical analysis covered three general areas: the to several groups of participants could have introduced bias demographic data gathered; the quantitative data gathered in the administration of the experiment, as the investigator’s in the first task; and the qualitative data gathered from the behavior might have varied from session to session. To min- second task. These results will be discussed later. imize this possibility, instructions for the experiment were scripted, and all groups thus received the same instructions. Overview of Demographic Data As a further means of guaranteeing impartiality, a third party was hired to administer the experiment. This person was Thirty-three males and 49 females participated. The partic- responsible for interacting with and instructing the partici- ipants ranged in age from 17 to 46, with a mean age of 20.21. pants in the experimental setting. To minimize experimental As might be expected of a sample drawn from an under- bias, this person was not informed of the research questions. graduate course, a large number of participants (67) were Additionally, the experiment administrator restricted her aged 17 to 20. Twenty-three participants (28%) had not yet interactions with the participants. She stood apart from the determined their majors. The remaining data regarding participants, so she could not see any particular participant’s majors were categorized broadly. Those that identified a progress, and only approached participants if they were hav- major tended to indicate majors in the applied sciences (28, ing trouble and specifically requested assistance. When this or 34.1%), social sciences (12, or 14.6%), or business (13, or occurred, the administrator answered only the particular 15.9%). The other categories were: humanities (3, or 3.7%); question that the participant posed, and then moved away. vocational (2, or 2.4%); and law (1, or 1.2%). Participants were seated in the labs, and the adminis- The participants claimed to use computers quite heavi- trator briefly explained the study, and said that all informa- ly: 44 (53.7%) said they used computers every day; 24 tion gathered during the test was to be kept confidential. (29.3%) used them 2 to 4 times a week; 9 (11.0%) used them Participants then answered the questions of the pretest once a week; and 5 (6.1%) used them once a month. The questionnaire. Participants were all given a tutorial session mean computer use level was 3.30, which translates approx- designed to help them familiarize themselves with the inter- imately to the 2 to 4 times a week category. Tests to see face, during which time the administrator addressed any dif- whether computer use differed between the factor groups ficulties participants had using the system. Then the found no significant differences(labels factor: χ2=1.45758, administrator presented to the participants the problem sit- p=.69210; content factor: χ2=1.95758, p=.58126). uation, reading from the problem statement. Participants In contrast, library online catalog use was much lower: selected the appropriate number of items for inclusion on 6 (7.3%) stated they used library catalogs 2 to 4 times a their bibliography, and then answered the qualitative por- week; 13 (15.9%) used them once a week; 34 (41.5%) used tion of the experiment. During the experiment, the admin- them once a month; and 29 (35.4%) had never used library istrator monitored the test site, to ensure that the online catalogs. Mean library online catalog use was .95, experiment proceeded smoothly. which translates to just below the once a month category. Once again, no significant differences were found between the groups for either of the factors. Results It was found that that most participants had little knowledge of the problem topic. Knowledge of big band The experiment was conducted in mid-October 1996. Eighty- music broke down as follows: 2 (2.4%) said they knew a lot two participants took part over a four-day period. The number about it; 24 (29.3%) knew a bit about it; 32 (39.0%) had of participants in each of the four interfaces was as follows: heard of it; and 24 (29.3%) knew nothing. The mean standard content, labeled display, 19 participants; enhanced knowledge level of big band music was 1.05, which equates content, unlabeled (ISBD) display, 19 participants; standard approximately to having heard about it. Knowledge of content, unlabeled display, 22 participants; and, enhanced Duke Ellington broke down as: 2 (2.4%) were quite famil- content, labeled display, 22 participants. Thus, when exam- iar with him; 16 (19.5%) knew a bit; 31 (37.8%) had heard ined by each of the two factors (absence or presence of labels, of him; and 33 (40.2%) knew nothing. The mean level of and standard or enhanced content brief screens), the total knowledge about Duke Ellington was .84, which again number of participants in each group was 41. equates approximately with having heard about him. No Data were compiled into a database, and output from significant differences were found for knowledge of Duke this database was analyzed using a statistical software pack- Ellington on the primary factors. 45(1) LRTS The Effect of Interface Design on Item Selection 33

There was no significant difference of knowledge of big eliminating this selection function; these screens differ in band music when compared on the content factor; however, that the full screen is usually more difficult to process there was a significant difference when compared on the because there is more information on-screen to examine. labels factor (χ2=8.45833, p=.03743). Because this was the Another important reason for examining each type of only significant difference found in the two topically orient- screen separately was that the experiment’s factors both ed variables, a second comparison was made between the relied on particular screen displays for the comparisons. The average of the two values in the topical questions and the screen content factor—that is, the effect that enhanced experiment factors. When grouped this way, no significant screen content had on participant behavior—relied solely on differences were found. the difference in brief display content for the different In summary, the participants in the experiment were groups. Similarly, the absence or presence of labels occurred more likely to be female than male, and generally in the tra- only on the single item brief and full display screens. Thus, ditional age range for college students. They were most like- it was decided to examine each type of screen separately, to ly to be majoring in the sciences, although a good portion test the effect these factors had on performance. had not yet decided on a major. They used computers Further analysis was then carried out using data com- approximately 2 to 4 times a week, but rarely used library piled for both the first 10 and the first 20 screens viewed. online catalogs. They had only a passing knowledge of the Because it was presumed that participants would become two aspects of the problem topic. The distribution of partic- increasingly familiar with the interfaces they used, and that ipants to the various experimental groups was not signifi- this learning process might manifest itself in behavior dif- cant. ferences over the course of the session, it was concluded that an overall measure of session activity might not give an Task 1: Creating a Select List accurate view of the ease with which participants used these Overview of the Data systems, and that separating the first screens each participant viewed might give a clearer sense of the learning For the first task in the experiment, a total of 5,970 log process. It was judged, moreover, that comparing among the entries were tallied. A problem with the logging of the first factors might also give an idea of whether one group of par- screen in 24 of the sessions required that these entries be ticipants learned to use the system faster than others, and dropped from the data set, resulting in a valid set of 5,946 that further analysis based on a comparison of the data from log entries. Help screens were also dropped, as they were the first 10 screens with those of the second 10 screens used so infrequently (27 times overall), that their inclusion might enable us to pinpoint the rate at which the learning affected statistical analysis. The total number of log entries process occurs. used for all further analysis thus was 5,919. Each log entry It must be noted that in compiling these initial screen included: the command issued, the record number of the statistics, erroneous durations for the first screen of 24 ses- item being viewed, the type of screen being viewed (Index, sions were dropped from the sample; consequently, the sta- Brief, Full, or Help), an error code, and the start and end tistics for these items includes data for screens beginning times for the screen. These data elements make it possible with the second screen viewed. to analyze the effects of the two factors on both the fre- Several more analyses were carried out on the duration quencies of screens viewed and on the duration of time data alone. For this experiment, the question of the speed spent on screens. with which participants performed the tasks presented to Within each data type category (i.e., frequencies and them was an important one. Speed in a test like this might durations), analysis was performed in several different steps. be measured in a number of different ways—for example, First, all log entries for each complete session were analyzed we can measure the overall time spent to perform the task in aggregate. Then, the data were separated by screen type or the mean time spent on specific screens. Another method to provide a more detailed method for testing the behavior for measuring speed is to use the number of characters of the participants while performing the task. processed per second to test participant performance. This Making this distinction was necessary for a number of measure can be derived by counting the total number of reasons. First, each of the three screen types used in the characters on each screen that the participant viewed and experiment had different content and complexity of data. dividing that by the duration of time spent on the screen. Index screens are columnar, making scanning easier, This measure controls for the amount of information that because like elements are grouped visually on-screen. participants viewed, and might help determine most accu- However, the inclusion of more than one bibliographic item rately the speed with which they were able to use the cata- complicates the user’s decision on which items to view in log information presented to them. greater detail. In contrast, both brief and full display screens Building on the idea that performance might change present information for just one bibliographic item, thus over time, analyses of screen durations for the first 5 screens 34 Thomas LRTS 45(1)

of each screen type were conducted. These were done based factors, the difference between groups on the label factor on the hypothesis that each screen type presents its own was not significant, while the difference on the content fac- processing problems to participants; analyzing each screen tor was significant (χ2=138.40842, p=.00000); participants separately might enable us to pinpoint to a greater degree using the enhanced content interface viewed significantly where users have difficulties with catalogs. more brief screens, and significantly fewer full screens. One final category of data was gathered in this portion Chi-squared tests on each of the demographic variables of the experiment: the participants’ final selections for their were found to be significant. Women viewed fewer index hypothetical reading lists. Analyses of these data were con- screens, fewer full screens, and more brief screens than did ducted to test whether the experimental factors or demo- men (χ2=62.28387, p=.00000). Infrequent users of comput- graphic variables had any effect on the items that ers viewed fewer index screens than did the heaviest users participants selected. (χ2=82.74381, p=.00000). Participants who said they had never used library catalogs, and those who used them week- Analysis of Frequencies of Screens ly viewed fewer full screens than others, while monthly users viewed more index screens (χ2=83.02628, p=.00000). Overall Frequencies. The mean number of screens Participants who knew nothing about big band music viewed in each session was 72.51, with a standard deviation viewed fewer full screens (χ2=100.41911, p=.00000). (SD) of 25.81 and a range from 30 to 177 screens viewed. Similarly, those who knew nothing about Duke Ellington Means for each of the groups in the two factors were com- viewed fewer full screens (χ2=32.52557, p=.00001). pared using a t-test, with no significant differences being Participants who had not decided on a major and those in observed. Analyses of mean number of screens viewed for the social sciences both viewed fewer full screens, while each type of screen were conducted. For index screens, the those in applied sciences viewed more full screens overall mean was 26.02 with a SD of 13.34 and a range from (χ2=182.65344, p=.00000). 1 to 58. Neither of the factor-based means varied from that First 10 Screens. Tests of the frequency of each noticeably. There was, however, a significant difference in screen type for the first 10 screens were conducted on both the mean number of index screens viewed by gender, with primary factors. No significant difference was found with males viewing on average 30.06 index screens and females the contents factor; a significant difference was found viewing 23.31 (t=2.23, p=.030). between groups on the labels factor. Participants using For brief screens, the overall mean number of screens labeled displays viewed fewer full screens and more index viewed was 25.51 with a SD of 20.35 and a range from 0 brief screens than did those using unlabeled displays screens viewed to 115. When examined by screen content, (χ2=11.09244, p=.00390). T-tests for each screen type were the means differed, with enhanced content participants view- conducted; a significant difference was found for the mean ing 29.46 brief screens compared to 21.56 screens viewed by number of full screens viewed per session when compared standard content participants. This difference was not signifi- on the labels factor (t=2.08, p=.040). cant at the .05 level, but was significant at the .10 level—thus In analyzing the first 10 screens compared to the demo- there is the likelihood that significant results would be found graphic data, a number of significant results were found. in further studies (t=-1.78, p=.08). When compared on the Chi-squared tests comparing screen type by level of com- labels factor, the difference was not significant. puter use yielded significant results (χ2=14.18658, For full screens, the overall mean was 20.65 with a SD p=.02762); participants with the least computer experience of 24.57 and a range from 0 to 146. When examined by did not use full screens at all, while other groups used them screen content, the means differed. For full screens, the in the neighborhood of 15% of the time. Novice users relationship of the numbers was reversed from that found instead viewed more brief screens than others. Significant with brief screens; enhanced content participants viewed on differences were also found when comparing library use to average 15.95 full screens, while standard content partici- frequency of screen type (χ2=20.16583, p=.00259). Those pants viewed 25.34 screens. Again, this difference was not with the most library experience viewed more full screens significant at the .05 level, but was significant at the .10 level than others. (t=1.75, p=.085). Differences between groups on the labels Differences among groups with regard to knowledge of factor were not significant. The number of help screens big band music were significant (χ2=16.10567, p=.01320), viewed was so low as to preclude analysis. while differences among groups based on Duke Ellington Chi-squared tests were also performed, comparing fre- knowledge were not. Those who had heard of big band quencies of the different screens viewed to the two factors, music viewed more brief and fewer full screens. Significant as well as to the following demographic variables: gender, differences were found between gender groups computer use, library use, big band knowledge, Duke (χ2=15.16189, p=.00051); men viewed more index screens Ellington knowledge, and major category. For the two main and fewer brief screens than did women. When examined 45(1) LRTS The Effect of Interface Design on Item Selection 35

by major, those in the social sciences viewed more brief tor differed, although this difference was not significant (t= screens than other groups (χ2=28.55680, p=.00458). -1.38, p=.172). Nor were the means significantly different First 20 Screens. The first twenty screens of each ses- when compared on the content factor (t=-.55, p=.583). No sion were analyzed separately, in order to isolate potential significant differences were found for overall mean per learning by the participants further. T-tests compare the screen duration among any of the demographic groups. mean number of each screen type viewed to each of the Next, each session’s data, separated by screen type, was primary factors as well as to each of the demographic vari- examined. For index screens, the mean was 11.90 seconds ables. No significant differences were found. Chi-squared with a SD of 4.85 and a range from 5.25 to 39.00 seconds. tests of the frequency of each screen type yielded several When examined by screen content, the means were 12.77 significant results, similar to those found for the first 10 and 11.04 seconds for standard and enhanced screens screens. For the labels factor, participants using labeled respectively, a difference that was not significant (t=1.63, screens viewed more index screens and fewer full record p=.108). Comparing by layout, the means were not found to screens than did those using the unlabeled screens (χ2=7.94, be significantly different (t=-.03, p=.977). When examined p=.01889). For the content factor, participants using the by library use some variation is seen, especially between enhanced content interfaces viewed more index and brief users with no experience (13.27 seconds) and once-a-month screens, and fewer full screens, than did those using the users (11.12 seconds), but these differences were not signif- standard displays (χ2=7.64, p=.02190). icant (f=1.31, p=.3134). Comparisons on other demograph- Chi-squared tests on demographic variables were all ic variables were not significant. It is interesting to note that significant. Men viewed fewer brief screens and more full the index screens were identical, regardless of the interface screens than did women (χ2=16.98349, p=.00021). Daily viewed. computer users viewed more index screens and fewer full For brief screens, the mean was 5.56 seconds with a SD screens than did others (χ2=30.85779, p=.00003). Those of 2.38 and a range from 0 to 13.76. No noticeable differ- who said they used library catalogs 2 to 4 times a week ences were found in the means for each factor; no significant viewed more full screens than did those with less catalog use differences were found in the means for demographic vari- (χ2=31.57513, p=.00002). Participants who knew nothing ables. For full screens, the mean duration was 7.70 seconds about big band music, or had only heard of it viewed fewer with a SD of 4.76 and a range from 0 to 22.00 seconds. No full screens than did those more familiar with the topic; in noticeable differences were found in the means for each pri- addition, those who had heard of it viewed more brief mary factor. When examined by level of computer use, a sig- screens than did others (χ2=30.45690, p=.00003). nificant difference was found (f=6.2535, p=.0007). Analyses Comparing Screens 1–10 with Screens 11–20. The on other demographic variables were negative. frequencies of screens viewed in the first 10 screens were First 10 Screens. The next group of data used for compared to the frequencies of the next 10 screens to deter- analysis was made up of data from the first 10 screens that mine whether there were any significant differences. each participant viewed. For the first 10 screens, the mean Paired-sample t-tests were conducted for each type of duration was 13.25 with a SD of 4.57 and a range from 3.30 screen; in each case, there were significant differences. to 30.90 seconds. When compared on the labels factor, the Participants viewed fewer index screens in the second 10 difference was not significant (t=-1.25, p=.214). The differ- screens than they did in the first 10 (t=5.03, p=.000), more ence between groups on the content factor was much small- brief screens (t=-2.31, p=.024), and more full screens (t= er and also not significant. When compared by gender, the -4.19, p=.000). differences were similar to those found for content. The mean durations for the first 10 screens varied more Analysis of Screen Durations when compared by computer use, and the differences were significant (f=4.4088, p=.0064). However, the means were Overall Duration. The first measure—overall time spent as follows: once a month, 14.02; once a week, 12.72; 2 to 4 to perform the task—is a broad measure that might provide times a week, 15.81; every day, 11.88. Thus, the means did a rough overview of relative performance. The mean session not seem to follow any pattern; perhaps this was a result of duration was 595.11 seconds, with a SD of 181.14, and a the limited numbers of participants in the lower experience range from 153 to 984 seconds. When examined by either of groups (5 and 9). A Scheffé test run on these variables iden- the main factors the overall means were virtually identical tified the difference between groups 3 and 4 as the source and not significantly different. of significance. On other demographic variables, no signifi- We can also examine the mean per screen duration for cant differences were found. each session. Overall, the mean per screen duration was When mean duration was examined by each screen 8.54 seconds with a SD of 2.34 and a range from 3.48 to type, there were no significant differences found when com- 18.92 seconds. Means between the groups on the labels fac- pared by content. Similarly, there were no significant differ- 36 Thomas LRTS 45(1)

ences in mean duration by each screen type when compared factors. Significant differences were found when comparing by labels. full screen duration to the computer use variable (f=6.9444, First 20 Screens. The first 20 screens of each session p=.0003). Otherwise, no significant differences were found. were isolated and analyzed. For the aggregate first 20 Duration by Screen Length. One final analysis was screens, there were no significant differences found when conducted using a measure based on the number of charac- comparing mean screen duration by content or by labels. ters viewed per second. These data were calculated by Once again, computer use yielded significant results measuring the number of characters on each screen viewed, between groups 3 and 4 (t=3.09, p=.0318). Analysis on other and dividing by the number of seconds spent on the screen. demographic variables yielded no significant results. Overall, the mean screen length was 531.055 characters, and Comparing each screen type for overall duration and mean the mean screen speed was 109.68 characters per second. screen duration yielded no significant results on either of The mean index screen speed was 167.97 characters per sec- the primary factors. Analyses of each screen type by the ond; the mean brief screen speed was 67.97 characters per demographic variables were not significant. second; the mean full screen speed was 87.75 characters per Comparing Screens 1–10 with Screens 11–20. second. Neither of the primary factor groups differed signif- When comparing data for screens 11–20 with screens 1–10, icantly when compared on the basis of these measures. we find some interesting differences. The mean for screens Similarly, most demographic variables yielded no significant 11–20 was 8.39 with a SD of 2.80 and a range from 3.20 to differences. When compared by gender, however, a signifi- 19.30. This represents a substantial drop in mean duration, cant difference was found; males processed more characters from 13.57 seconds. In addition, the standard deviation per second than did females (t=2.04, p=.045). dropped substantially, from 4.57 to 2.80, which indicates that the variance also dropped substantially. When using a t-test Analysis of Selection Data for paired samples, the difference in mean times was found to be significant (t=11.79, p=.000). Moreover, the ranges and Data regarding the items that each participant selected were histograms in each grouping suggest that the learning compiled and examined. Of particular interest in this area process affected primarily the longest times in the range. was the type of screen from which the participants made Mean durations for each screen type for screens 11–20 their selections. Overall, the participants made 809 selections were examined. No significant differences were found when out of a possible 820; one participant selected only 8 items, comparing index screens on the primary factors. Brief while a second selected only 1 item. Participants selected 71 screen mean duration differed for participants viewing stan- items directly from index displays, 390 from brief record dis- dard or enhanced screens; participants viewing enhanced plays, and 348 from full record displays. When chi-squared screens spent longer on these screens (enhanced=5.90, stan- tests were conducted against the two factors, a significant dif- dard=4.78). This difference was not significant at the .05 ference was found between groups on the content factor level, but was significant at the .10 level, suggesting that (χ2=8.93966, p=.01145). Participants using the enhanced there is a likelihood that significant results would be found interface made fewer selections from index screens and full in a study designed specifically to test for this (t=-1.93, screens, and more selections from brief screens. p=.058). Full screen mean duration was significantly differ- No significant differences were found with the labels ent when compared by content (standard=10.44, factor. Significant differences were found between groups enhanced=8.16) (t=2.02, p=.043). Participants using the on computer use (χ2=26.31462, p=.00019) and library use normal screen interface took longer than those using the (χ2=55.07852, p=.00000). On both measures, novice users enhanced interfaces. There were no significant differences were more likely to select items directly from index screens found when compared by the labels factor. than experienced users. On computer use, experienced First 5 Screens of Each Type. The first 5 screens of users were also less likely than all other groups to use index each type were separated from the main body of data in screens to make selections. On library use, the most experi- order to isolate the performance differences that partici- enced participants were more likely to use full screens to pants might have had in relation to the different screens. make their selections than were all other groups. The mean duration of time spent on all of these screens was Tests on gender, knowledge of big band music, and tested for significance against each of the factors and the knowledge of Duke Ellington were not conclusive. demographic variables. There were no significant differences when the primary factors were examined. Significant Task 2: Judging Relevance differences were found between daily users of computers Overview and those using them 2–4 times a week (f=4.9600, p= .0034). When examining each type of screen separately, no The second task in the experiment required the participants significantly different results were obtained on the primary to provide basic relevance ratings for 20 items on the same 45(1) LRTS The Effect of Interface Design on Item Selection 37

topic as in the first task, along with an indication of the fields categories were separate, no significant results were that they used to make their relevance judgment. In this obtained when these categories were conflated. task, participants were presented with either a full or brief record in the display mode for their session (i.e., labeled or Analysis of Fields Checked not, enhanced content or not), and were asked to decide The final group of data to be analyzed in the second task whether they would choose to examine this item further. consisted of the check rates for each of the fields that par- Participants could choose: ticipants viewed. Participants checked 2,058 data elements as being useful to their decision, a rate of 1.48 items per 1. that they would examine this item on the shelf; selection. The rates at which participants checked particular 2. that they would NOT examine this item on the shelf; fields as useful ranged from a high of 70.73% for the sum- or, mary field (MARC 520) to 0% for several fields. The title 3. that they could not decide from the information pro- field had the second highest check rate at 60.55%, followed vided. by the subjects field at 50.8%. The next group of fields was checked at a much lower rate, and consisted of the series If they chose either 1 or 2, they were required to indi- (13.72%) and general notes (12.56%) fields. Four additional cate which data elements enabled them to make their deci- fields were checked at rates near 5%, and 5 more at rates sion, by checking boxes next to the data elements that they around 1.5%. Four fields were not present in the sample. used to make their decision. One thousand six hundred and nineteen (78.67%) of the The data for this task thus consist of two broad cate- checks were made by participants in one of two fields: the title gories: the relevance rating, and the fields used in decision field (MARC 245), and the subjects field (MARC 6xx). making. The relevance rating was required, so each session Overall, significant differences between factors were found for had a total of 20 relevance ratings. For the fields used in the the following MARC fields: 300, 4xx, 6xx, and 090. Field 300 decision-making process, it was necessary (for statistical pur- was significant (χ2=30.35, p=.000) in large part because of an poses) to determine the total frequency that each field was extremely high check rate for the standard labeled display; sec- viewed by each participant. This was a result of the variabil- ondarily, a low check rate in the standard unlabeled interface ity of MARC records. This frequency count differed for also contributed. Field 4xx was significant (χ2=17.30, p=.001) each interface, because the brief screens differed. For all primarily because there were no checks at all in this field for the data gathered in this task, chi-squared tests were used. the enhanced unlabeled interface. Field 6xx was significant (χ2=10.86, p=.002) because of a high check rate in the Analysis of Ratings enhanced labeled display and a low check rate in the standard unlabeled display. Field 090 was significant (χ2=10.81, p=.002) Participants viewed a total of 1,640 bibliographic records in because of variance found in all 4 interfaces. This result might either brief or full display. They checked 999 items (60.91%) not be reliable because of low overall frequencies. as being useful, 390 items (23.78%) as not useful, and 251 When examined by demographic variables, many simi- items (15.30%) as not providing enough information to make lar results were obtained. Using computer use as the group- a decision. The differences in frequency between groups on ing variable, significant results were found for the Not the content factor were significant (χ2=48.54232, p=.00000). Useful and Don’t Know categories, and fields 300 and 6xx. These results occurred in large part because of the high num- Using library catalog use as the grouping variable, significant ber of Not Useful entries by the enhanced, labeled interface, results were found for categories Not Useful and Don’t as well as a combination of high and low counts among the Know, and fields 245, 300, 6xx. In this instance, field 245 was groups in the Don’t Know category. What is more important significant in large part because of the low check rates by the is that the high frequencies for this category were observed in most experienced library catalog users. When using knowl- the standard content interfaces, while the low frequencies edge of big band music, significant results were found for were observed in the enhanced content interfaces. the Not Useful category, and fields 245, 300, and 6xx. Field Further analysis of the data revealed that the ratings 245 was significant here because of a high check rate by were linked to the type of screen that was viewed. When those with no knowledge of big band music. Significant dif- comparing the screen type to the relevance rating, a signifi- ferences were also found for fields 020 and 090, but the indi- cant difference was found (χ2=60.21908, p=.00000). This vidual cell counts were too low for conclusions to be drawn. difference was due to the effect of the high rate of Don’t Know classifications for brief screens. Summary of Statistical Analysis Secondary analysis was conducted on the data set, in which the Don’t Know category was combined with the Not The statistical analysis was presented in two broad cate- Useful category. Contrary to results received when these gories: analysis of data from the first task, and analysis of 38 Thomas LRTS 45(1)

data from the second. For the first task, participants using with enhanced content brief screens viewed more brief enhanced brief screen interfaces viewed more brief screens screens and fewer full screens overall than did their coun- and fewer full screens than did their counterparts. Screen terparts. The enhanced content of the brief display enabled durations for the second 10 screens were found to have participants to avoid viewing full displays. Given that full dropped from those of the first 10 screens. Statistical analy- displays contain more data, more data elements, and by all ses comparing demographic variables to the screen frequen- measures require more time to process, a reduction in the cies uncovered many significant differences. Participants number of full display screens viewed most likely represent- using the enhanced content interfaces made fewer selec- ed a reduction in the effort that users expended to achieve tions from index and full screens, and more selections from their goals. Thus, the null hypothesis in this case was reject- brief screens. For the second task, participants who used ed; the number of screens viewed at each level of specifici- enhanced content interfaces were able to make some sort of ty did differ based on whether participants used enhanced relevance judgment more frequently than were those who content brief displays or not. used standard content interfaces. In the next section, the It should be noted that these differences were not importance of these results will be discussed, with specific observed when the first 10 screens were examined in isola- regard to the research questions and hypotheses. tion. This finding suggests that the participants altered their behavior during the course of the test. That behavior had changed by the twentieth screen, however, as the frequency Discussion data for the first 20 screens already show a significant dif- Research Question 1: Screen Content ference between the groups based on-screen content. It is also interesting to note that participants using The first research question asked whether there was a cor- labeled displays viewed more index screens and fewer full relation between screen content and the time it takes a par- screens for the first 10 and first 20 screens than did those ticipant to perform a task. Three hypotheses were derived using unlabeled displays. This contradicts results for the from this broader question. The first was that the number of labels factor that were obtained from the overall data, where screens viewed would differ between treatment groups. The there were no significant differences. This suggests, in turn, second was that the number of screens viewed at each level that users of labeled interfaces initially preferred viewing of specificity would differ between treatment groups. The index screens to single record displays. That the absence or third was that the screens used to make selections decisions presence of labels would affect a participant’s screen selec- would differ between treatment groups. tion—even for a short period of time—was not an anticipat- ed outcome. Hypothesis 1.1 The behavior change noted above is also reflected when both the frequencies and the durations of the first 10 In this experiment, screen content was varied between screens are compared to the next 10 screens. The partici- treatment groups so that some groups viewed more topical- pants viewed fewer index screens and more brief and full ly oriented data on the brief screen than did other groups. screens in the second 10 screens than they did in the first 10. Frequency data were tabulated for all participants in sever- This might reflect the fact that many participants appeared al categories—by the main factors in the experiment, by the to use the first few screens to familiarize themselves with the demographic variables, and broken down by type of data set and with the system. Or it might be that they found screen—and statistical tests were run. There were no signif- the brief and full screens more informative for their needs. icant differences between the groups in the overall number One way or the other, a second change in activity is seen in of screens that participants viewed on either the content fac- the fact that mean screen duration dropped a significant tor or the labels factor. Thus the null hypothesis, which amount in all types of screen, as well as for all screen types states that there is no difference in the number of screens taken together. These drops suggest that the participants viewed, cannot be rejected. Therefore, the first hypothe- became familiar with the system, its content, and how to use sis—that there will be a difference in the overall number of both to solve the problem at a relatively rapid rate. screens viewed—cannot be supported. It is possible that these drops—both statistically significant and substantial—are the result of some other factor, Hypothesis 1.2 such as that the participants became dissatisfied with the system and simply hurried through the task to be done. When considering the second hypothesis for this question— There is some evidence in the post-test comments that sug- i.e., whether there were differences in the number of indi- gest that this might be the case, inasmuch as some partici- vidual screen types examined—significant differences were pants commented that there was not enough information in found. It was found that participants who used the interfaces the system for them to make the types of decision that were 45(1) LRTS The Effect of Interface Design on Item Selection 39

requested of them. However, it would appear from the of different steps to follow up on this state of indecision— transaction logs that the participants figured out their pre- request more information, browse the shelf, etc. It is quite ferred method of solving the problem within the first 20 or possible that these two categories cannot be conflated. 30 screens, and then used that method with increasing efficiency as the problem progressed. The participants seemed Demographic Effects to develop routines that they followed throughout the experiment—for example, one participant would view an item in Further complicating the findings of this study are the signif- the brief display, then the long display, and finally select that icant results obtained on many of the demographic variables. item. It is assumed that in adopting such a routine, users can Most noteworthy of these results are the consistent significant work more quickly because they gain practice and experi- differences found when examining the results using comput- ence in carrying out the steps in the task. er use as a grouping variable. More experienced computer users generally viewed fewer detailed screens. This might be Hypothesis 1.3 explained by the fact that these users were less likely to accept the system limitations—both in terms of content and func- Because the problem presented to the participants was sub- tion—than less experienced users. Computer users have ject-oriented in nature, it was thought that the use of become more demanding as the capabilities of computers enhanced content brief displays would have an effect on the have grown, and it seems reasonable to imagine that they selection behavior of the participants. The third hypothesis would be demanding in this experiment. Several participants for research question 1 asked whether the screens from commented on the system limitations; for example, some which participants made selections would differ based on complained that there was not enough information in the sys- the experimental factors. Specifically, it was thought that tem to make a decision, while others complained at having to those who used an enhanced content interface would be use the keyboard to issue commands. more likely to make their decisions from brief record dis- In addition, gender had a significant effect on several of plays. Statistical analysis of the selections that participants the measures. The most interesting of these was the differ- made shows that this in fact occurred—users of the ence that was found in the frequency that males and females enhanced content interfaces selected 217 items from brief viewed different screen types. Women viewed fewer index displays while their counterparts selected only 173. screens and more brief screens than did men. These results Confounding this is the fact that users of standard content bring to mind the study conducted by Ford and Ford interfaces selected a larger number of items from index (1993), where the researchers found that experiment partic- screens. This might have been skewed by the actions of the ipants used three broad search strategies that were associat- one or two participants who made all their selections from ed with one gender or the other. The results of this study index screens. Despite this anomaly, the null hypothesis was serve to confirm those findings, because women in this rejected, and the hypothesis confirmed; the screens used to experiment tended to view records at a greater level of detail make selection decisions did differ between treatment than did men. groups. The third variable with noticeable effect on the results Additional support for this conclusion can be found in was the library use variable. Those with higher library cata- the data from the second task, where participants were log experience were more likely to view full display screens asked to make basic relevance judgments for 20 items pre- than those with less library catalog experience. This suggests sented to them either in brief or full display. Users of the that experienced library users are more likely to exploit the enhanced content interfaces checked that they could not resources in a catalog more fully, because they are more decide the value of an item significantly fewer numbers of likely to examine the fuller information. Finally, knowledge times than users of the standard interfaces. The decrease in of the topic was linked to the likelihood that the participant the Don’t Know category was largely transferred to the Not would view fuller information. Useful category, which suggests that the increased informa- It must be pointed out that tests to determine whether tion helped participants make negative selection decisions. the main treatment groups differed on any of the demo- It should be noted that the secondary analysis—in graphic variables recorded were negative in all cases except which the Don’t Know and Not Useful categories were com- in the instance of the knowledge of big band music. It was bined—confounds these conclusions to a degree. These cat- concluded, therefore, that these demographic variables did egories were combined under the assumption that a user’s not have a significant or systematic effect on the tests of the indecision would most likely translate into a decision not to primary factors. With regard to the knowledge of big band examine an item further, and the results were not signifi- music variable, only two participants said they were familiar cant. However, this assumption might not be valid; in an with big band music, which makes it unlikely that their active environment, the user might choose any of a number actions would adversely affect the validity of the results. 40 Thomas LRTS 45(1)

Research Question 2: Screen Layout as a measure, it is quite possible that participants only processed part of the information on-screen. Thus, a partic- Research question 2 asked whether the layout of biblio- ipant might only have read the title and subject data ele- graphic data on-screen would affect the speed with which a ments on a given screen, rendering the total screen measure participant could extract information from it. Two hypothe- less meaningful. ses were derived from this broader question. The first was that the time needed to complete the task would differ Demographic Variables between treatment groups. The second was that the time that participants spend on specific screen types would differ While layout of the data on-screen had no measurable effect between treatment groups. on the ability of participants to perform the task presented For this experiment, layout was defined as the absence to them, other factors did. The level of computer use affect- or presence of on-screen labels delineating the data ele- ed participant performance in the first 10 screens, the first ments of the bibliographic record. The data gathered that 20 screens, and the first 5 screens of each type, although address this question consist of the transaction logs and the these results are clouded by the fact that the differences in screen duration information derived from those logs. While each case failed to follow a single progression. This might previous researchers have found that labeled displays are have been due to the uneven distribution of individuals over easier for users to read (Tullis 1983; 1988), no such assump- the computer experience variable, as there were so few par- tion was made for this experiment; thus, the t-tests that were ticipants who had little experience with computers. performed were two-tailed. Consequently, it is not clear how these data might be interpreted. The complex nature of the values for mean screen Hypothesis 2.1 duration here suggest the possibility that some other variable interacted to create these results. To test whether screen layout had an effect on the time it However, even this difference disappeared when the took to perform the task, the data were examined from sev- length of individual screens was introduced as a control. eral different perspectives. First, overall session duration Using this more exacting measure of speed, the only signifi- was compared against the labels factor. This test failed to cant difference was found on the gender variable—males reveal significant differences. Second, the mean durations processed more characters per second than did females. for all screens in each session were compared, with no sig- This result might have been due in part to the fact that nificant differences observed. Additional analyses of the males viewed significantly more index screens than did effect of labels on-screen duration included: mean duration females. Index screens had a higher mean processing speed of the first 10 screens viewed and mean duration of the first than either of the other two screens; higher levels of index 20 screens. No significant differences were found. Thus, the screen viewing would thus result in higher overall process- null hypothesis cannot be rejected, and the hypothesis that ing speed. screen layout had an effect on participant performance can- Based on the data gathered in this experiment, there is not be supported. no evidence that screen layout has an effect on user performance. Neither of the two hypotheses under this ques- Hypothesis 2.2 tion were supported, regardless of the method of measuring this difference. This conclusion is remarkable, given the To test whether screen layout had an effect on the time that body of human factors research that has been conducted participants spent on particular screen types, a number of that contradicts this (c.f., esp., Tullis 1981). tests were conducted. First, the overall mean screen dura- One possible explanation for these results is that it is tion for each screen type was compared, with no significant possible that the design of the experiment was not sophisti- differences found. Additional analyses of the effect of labels cated enough to obtain wholly reliable results; for example, on-screen duration included: mean duration of the first 10 it is possible that some participants just selected the first screens viewed, separated by screen type; mean duration of records they viewed simply to be finished with the test. the first 20 screens by screen type; mean duration of the first Examination of the transactions logs, however, suggests that 5 screens of each screen type; and, the number of characters most participants made at least a modest effort at carrying processed per second. In all these tests, no significant dif- out the task. For example, while the frequency distribution ferences were found. Thus, the null hypothesis here cannot of screens viewed for each record in the set shows a general be rejected; screen layout had no effect on-screen durations decreasing trend as the record number increases, there still for particular screen types. remains a wide dispersion of activity. With regard to the characters per second measure, it If large numbers of participants failed to give some should be noted that while overall screen density was used effort to the experiment, it seems reasonable to expect that 45(1) LRTS The Effect of Interface Design on Item Selection 41

extremely high view rates for the first and last records in the participants made were placed in 2 fields: the title field list would have been observed—those closest to the starting (MARC 245) and subjects field (6xx). point of the task. This was not the case. Even if it were, the This represents an even higher concentration of inter- fact that some participants chose not to expend a great deal est than found in those earlier studies, but might be of energy trying to select items from the display does not explained by a couple of factors. First, the problem in this necessarily invalidate the results—especially given the fact experiment was highly subject-oriented; participants were that users often do not seem to be extremely diligent when asked to select items on a given topic. This might have examining catalog search results (e.g., Millsap and Ferl caused the participants to focus on those fields that had the 1993). most topical content. This can be further supported by the Another explanation for the lack of significant results in fact that the summary field (MARC 520), while not present this area might be that beyond a certain point the complex- in many records (82 chances overall), was selected by par- ity of layout has no effect. Bibliographic displays include a ticipants an extremely high percentage of the time that it large amount of data arranged in a complex set of data ele- was present (70.73%). ments, the combination of which can present a great deal of Second, the nature of the experiment limited the information to the user. In this experiment, the mean display chances that participants had to select other fields. The ran- length was 531 characters out of a possible 910—a 58% domly selected records and randomly selected display screen density level. Most researchers recommend screen screens emphasized primary fields and downplayed second- densities more in the neighborhood of 30 to 40%, a much ary ones. The title field was on every screen presented to the lower figure. Added to this is the fact that many of the data participants, and the subjects field in most. In contrast, the elements in a bibliographic display are neither clearly contents field—among other potentially useful fields—was explained to users nor inherently known by them, minimiz- not viewed at all. This uneven representation probably ing the effect that the absence or presence of labels might affected the relevance indications of participants. have. In this experiment, every attempt was made to use A third reason for this concentration might have to do clear and unambiguous terms for the labels; however, it is with the fact that the experiment was a simulation. One of not clear to what extent the participants knew and under- the fields that previous researchers said that library users stood the meaning of the information presented to them. found important was the call number field. Obviously, with- These complexities could have overwhelmed the effect that out the call number, locating the actual item on the shelf is labels might have had on participant performance. greatly complicated. In this experiment, however, participants were not asked to retrieve the items they selected; Research Question 3: Relative Value of thus, their perceived need for the call number may well Data Elements to Relevance Judgments have been mitigated. It is perhaps instructive to note that some participants selected call number despite the nature of Research question 3 asked whether there were particular the experiment. This might reflect an effort on the part of data elements that participants used more often than others those participants to take the simulation seriously, or it to make relevance decisions in topically oriented retrieval might be a retrieval strategy that they use. tasks. Two hypotheses were derived from this broader ques- The check rates of four other fields differed from the tion. The first was that those data elements most associated results of earlier research. Both the author and publication with topical bibliographic data would be selected most fre- fields in this experiment were selected as useful only a small quently by participants of the assigned task. The second was percentage of the time—at rates only half that found for the that the fields deemed useful would differ between treat- series field. On the other hand, the series and general notes ment groups. fields were fields that were not identified as important in earlier research, but that were checked by participants at a Hypothesis 3.1 low but steady rate. It is unclear why these results were obtained. One purpose behind including this question was to support The evidence regarding useful fields that was gathered or refute past research that identified the fields considered in this experiment seems to suggest that novice library users helpful by library catalog users—notably Seal (1983), consider only a few fields to be helpful to their selections Hufford (1991), and Palmer (1972). Those researchers decisions, and ignore the rest. In addition to the explicit found that the substantial majority of users considered just a check rates in the second task are the transaction logs in the few fields to be necessary or useful for their needs. In this first task, which indicated that novice library catalog users experiment, this hypothesis relied on the subjective feed- viewed more index screens and fewer detailed screens than back provided by the participants in the second task of the more experienced library catalog users. One possible expla- experiment. More than three quarters of the checks that nation for this behavior is that novice users, unfamiliar with 42 Thomas LRTS 45(1)

the data elements in a bibliographic record, focused on the ance—hypothesis 2.1 and 2.2 were both soundly rejected. simplest displays, where the data were presented in the For research question 3—whether some data elements are most uniform and unambiguous manner. Detailed screens more commonly considered useful than others—hypothesis can contain any of a wide variety of data elements, many of 3.1 was supported, while the results for the hypothesis 3.2 which are identified by terminology that is unclear to the are more ambiguous. Participants in the treatment groups user (e.g., “LCCN,” “ISBN,” “Music Number,” etc.). did select some data fields as useful at different rates; how- Avoiding these screens might reduce the user’s uncertainty ever, these could at best be considered tentative results. and unease in using the system.

Hypothesis 3.2 Conclusions

With regard to whether the fields deemed useful differed This research began with a desire to contribute to the between treatment groups, the significant results obtained knowledge about online catalog use. The primary objective for several of the fields partially support the hypothesis. was to explore the effect that layout and content might have When these results are considered in relation to the core on a typical information retrieval task. An experiment was fields identified by earlier research, it was found that both designed to isolate these aspects of online catalogs and the the subjects and call number fields differed significantly experiment was run with participants at the University of between treatment groups. Looking first at the call number Pittsburgh. The data gathered in this experiment were ana- results, it was noted that the check rates in each interface lyzed and the results were presented. diverged from the expected rates, and that these discrepan- In this research, several important discoveries have cies crossed both treatment factors. It might be that the low been made. First, it was found that alterations to the content overall check rate makes these data unreliable in this con- of particular screens—in this case, enhancing the content of text. the brief display—had a significant effect on the behavior of Turning to the check rates in the subjects field, partici- the experiment participants. Participants who used brief dis- pants using the enhanced content labeled display selected play screens that contained more topically oriented data ele- subjects at an especially high rate, while users of the unla- ments resorted to full display screens significantly fewer beled standard content display selected this field fewer times than did those using standard, citation-oriented brief times than expected. Finally, with both remaining significant displays. This finding is important because it suggests that— results, it is interesting to note that the low check rates in for topically oriented tasks, at least—brief catalog displays unlabeled displays were primary contributors to the result. might be redesigned to include more fields with subject-rich In addition, a high number of checks in the labeled display content (e.g., MARC 520 and 6xx fields). Such a redesign contributed to the results for the physical description field. would presumably reduce both the number of screens that Taken together, the significant results for subjects, users would need to view and the complexity of those physical description, and series fields follow a pattern: par- screens. This might, in turn, simplify the user’s task of find- ticipants who did not use interfaces with identifying labels ing wanted items in the catalog. indicated that they found these data less useful than those A second important finding of this research was that the viewing labeled displays. While these results do not form a layout of information on-screen had little effect on the time compelling argument to support the use of labeled displays it took experiment participants to complete the assigned in all cases, they certainly suggest that in some cases labels task. Whether measured as overall time spent to perform affect user perceptions of the data they see. Thus, while the the task or as average time spent on particular screens, no absence or presence of labels did not have a significant significant differences between treatment groups were effect on the speed with which participants performed the found. When combined with the fact that mean screen first experiment task, the results in the second task still duration dropped substantially over the course of the first 20 might affirm their importance to end users. screens, it seems likely that the participants of the experiment adapted to the mode of information presentation rap- Summary of Discussion idly, and became equally comfortable with that mode of presentation—regardless of the specifics of the mode of This discussion of the statistical tests included an analysis presentation. These results contradict earlier related and comparison to the research questions and hypotheses. research, and raise questions about whether bibliographic On research question 1—whether screen content affected data somehow differ from the data used in that earlier user behavior—hypothesis 1.1 was rejected, while hypothe- research. ses 1.2 and 1.3 were confirmed. With respect to research A third important finding was the support for earlier question 2—whether screen layout affected user perform- research regarding the fields that participants felt were use- 45(1) LRTS The Effect of Interface Design on Item Selection 43

ful to their judgments of relevance. Authors of earlier studies duration of screen measure meaningless in that particular had found that most users consider only a few fields impor- experimental context. tant in the bibliographic record. The results of this study con- The second method of conducting the experiment used firm these findings, and take those results further with regard a custom-made Visual Basic application, and was generally to topically oriented catalog tasks. The participants in this more reliable in performance. There were some problems, study overwhelmingly considered three MARC fields to be however. Whereas the first task relied exclusively on key- useful to their judgments of relevance—a fewer number of board input, the second relied on mouse actions, and this fields than was found in earlier studies. This discovery is change in system modes caused some confusion among the important because it might be used by online catalog design- participants—even after the instructions had been modified ers to select fields for inclusion on given catalog screens. This to emphasize this change. would give users more of the data that they feel are useful, without cluttering screens with data they feel are not. Future Research

Method Considerations A great deal of further research still needs to be conducted on the effect that layout of data on-screen has on user per- During the course of this research, two broad methods of formance. While the research outlined here showed that no conducting the experiment were tried. The first of these differences were found in performance based on the employed HTML and the Hypertext Transfer Protocol as absence or presence of data labels, many other tests could the mechanism for carrying out the experiment. Several and should be conducted. For example, the effect that over- observations can be made with regard to this system. First, all screen density has on usability should be explored with although there were very few problems with participants respect to bibliographic data. In fact, many studies that have using a mouse to navigate the system, one problem that was been conducted with other types of information systems observed was that many had trouble with bibliographic (e.g., Kruk and Muter 1984; Marcus 1982; Trollip and Sales records that were long enough to require scrolling. They did 1986; Tullis 1981, 1983) should be replicated in the special not know how to view the bottom of the record. It is unclear context of bibliographic data. why these participants had these troubles, but that these A number of other areas for further research have been troubles existed should serve as a point of consideration for uncovered. One such area is the effect that user knowledge others doing such tests. of bibliographic data has on user behavior. In the current A second observation is that data-gathering from partic- study, participants were asked to identify the data elements ipants in the HTML environment requires access to server- that they used to make basic relevance judgments. Left side processing. Without server-side control of the flow of unanswered is the question of how the individual partici- the experiment, participants can ignore requests for infor- pant’s knowledge of bibliographic data content might have mation posed to them. In the pilot test, 40% of the requests affected the results. It seems reasonable to assume that for participant feedback were ignored. While this response users who know more about the content and structure of an rate was due in part to the design of the particular experi- information system will be better able to exploit the infor- ment, a good portion no doubt resulted from the inability to mation in that system. Indeed, Brunner et al. (1992) found enforce compliance. Use of a dedicated HTTP server with some evidence to this effect. Similarly, the selection of data access to server-side control would have enabled the pilot elements used in a basic relevance judgment might depend test to have more success in this area; with the advance in on the user’s knowledge of bibliographic data structures. technologies, this approach would be more feasible now. Research must be conducted in which the effects that this Finally, extreme care must be taken in any such test to knowledge has on user behavior and on user relevance judg- ensure that the variable being tested is not confounded by ments are explored. Some research into the effect that user other related variables. In designing the pilot test, the vari- knowledge—both of topic and of system function—can have ables were tested simultaneously and interacted in such a on user behavior has already been conducted (e.g., Hsieh- way as to make reliable measure practically impossible. It Yee 1993), but additional research focused directly on layout was assumed that participants would proceed sequentially and content needs to be conducted. through the screens presented to them, using one screen to Linked to this study of users’ knowledge of bibliograph- examine the bibliographic information, and the next to pro- ic data structures would be research into the labels that are vide feedback regarding the data elements used for their used in online catalogs. Bibliographic data contain a large relevance judgment. In fact, the participants became famil- number of data elements. The labels that have been select- iar with the interface and often appeared to anticipate sev- ed for these data elements have not been tested with users eral screens in advance. This enabled them to ignore to see how clearly they communicate content and meaning certain screens and focus on others, thus rendering the to users. Research into the effect that different labels have 44 Thomas LRTS 45(1)

on user comprehension would enable catalog designers to types of fields that users utilize in given retrieval situations, use terminology with the widest recognition among users. this process can be accelerated, and indexes can be designed Another area of research that needs further exploration that truly match the user’s information need. is determining the data elements that are most important for It is necessary for additional research to examine the particular types of retrieval situations. This research would relationship between gender and online behavior from a have to start with a greater consideration of the types of number of perspectives. The current study seems to corrob- tasks users typically come to the online catalog to solve. orate the findings of Ford and Ford (1993), who found While it might be obvious that many users come to the cat- behavioral differences between the genders in online text alog to find out what a library has on its shelves, other uses use. Studies that specifically test the effect that gender has are evident, if less well defined (e.g., finding all the editions on the types of bibliographic information deemed useful in of a particular work, in order to identify the most authorita- given situations must also be conducted. tive edition). Anecdotal evidence regarding the uses to which catalogs are put has long been offered; what is need- Works Cited ed are studies that offer substantive, empirical data regard- Akeroyd, J. 1990. Information seeking in online catalogues. Journal ing these uses. Such knowledge is necessary before the of Documentation 46: 33–52. library and information science field can act to improve cat- Allen, B. 1994. Cognitive abilities and information system usabili- alogs in a meaningful and purposive manner. ty. Information Processing & Management 30: 177–91. Once a basic topography of the types of retrieval situa- Allen, B., and G. Allen. 1993. Cognitive abilities of academic librar- tions and their relative frequencies has been achieved, ians and their patrons. College & Research Libraries 54: research into the data needs for each of these situations 67–73. should be conducted. Common sense would suggest that Aykin, N. M., and T. Aykin. 1991. Individual differences in human- certain data elements would be more useful in some situa- computer interaction. Computers and Industrial Engineering tions than in others, and the research described here con- 20: 373–79. firms that assumption in one particular instance. Moreover, Bates, M. J. 1986. Subject access in online catalogs: A design it might be that the usefulness of data elements follows the model. Journal of the American Society for Information Science 37: 357–76. same skewed patterns found in other areas of library and ———. 1989. Rethinking subject cataloging in the online environ- information science research. Using the frequencies of ment. Library Resources & Technical Services 33: 400–12. occurrence as a means of setting priorities, library and infor- Beheshti, Jamshid. 1992. Browsing through public access catalogs. mation science professionals can systematically begin to Information Technology & Libraries 11: 220–28. evaluate the typical data needs in each retrieval situation. Belkin, N. J., P. G. Marchetti, and C. Cool. 1993. Braque: Design Such research would make it possible to design online cata- of an interface to support user interaction in information logs with basic screens that more closely reflect the needs of retrieval. Information Processing & Management 29: 325–44. the user in a particular retrieval situation. Belkin, N. J., and A. Vickery. 1985. Precursors to information seek- Research in this area might also be broadened to con- ing behavior: “Information need.” In Interaction in informa- sider such basic display issues as the order of display of data tion systems. London: British Library. elements on-screen. While online catalog displays have Benbasat, I., and P. Todd. 1993. An experimental investigation of interface design alternatives: Icon vs. text and direct manipu- taken as their model the catalog card—a familiar, comfort- lation vs. menus. International Journal of Man-machine able display—it is certainly possible that a different ordering Studies 38: 369–402. of data elements on-screen might be more quickly and eas- Borgman, C. L. 1986. Why are online catalogs hard to use? Lessons ily processed. Research should be conducted to determine learned from information retrieval studies. Journal of the the effects that data order might have on user performance. American Society for Information Science 37: 387–400. It might be possible to use the results of these studies as ———. 1988. Human factors in the use of information systems: a basis for redesign of online catalogs not just from a display Research methods and results. In Information research: perspective, but also from a functional perspective. With a Research methods in library and information science, edited clearer understanding of the types of retrieval problems with by Neva Tudor-Silovic and Ivan Mihel. London: Taylor- which users approach online catalogs, indexes and search Graham. modes could be reorganized to match those problems. For Brajnik, G., G. Guida, and C. Tasso. 1987. User modelling in intelligent information retrieval. Information Processing & example, instead of indexing a bibliographic database using Management 23: 305–20. MARC fields as the focus, the indexing might be designed so Brown, C. M. 1988. Human-computer interface design guidelines. that related data elements are joined in one index. To some Norwood, N.J.: Ablex. extent, library systems designers have begun to create such Brunner, H. et al. 1992. An assessment of written/interactive dia- indexes, as when they create keyword indexes that combine logue for information retrieval applications. Human- title and subject fields. With a better understanding of the Computer Interaction 7: 197–249. 45(1) LRTS The Effect of Interface Design on Item Selection 45

Buckland, M. K. 1991. Information and information systems. New Hufford, J. R. 1991. Elements of the bibliographic record used by York: Praeger. reference staff members at three ARL academic libraries. Buckland, M. K., B. A. Norgard, and C. Plaunt. 1993. Filing, fil- College and Research Libraries 52: 54–63. tering, and the first few found. Information Technology & International Conference on Cataloging Principles. 1963. Report: Libraries 12: 311–19. International Conference on Cataloging Principles, Paris, Cochrane, P. A., and K. Markey. 1983. Catalog use studies—since 9th–18th October, 1961. London: Organizing Committee of the introduction of online interactive catalogs: Impact on the International Conference on Cataloging Principles. design for subject access. Library and Information Science Kerr, S. T. 1990. Wayfinding in an electronic database: The relative Research 5: 337–63. importance of navigational cues vs. mental models. Coll, J. H., R. Coll, and R. Nandavar. 1993. Attending to cognitive Information Processing & Management 26: 511–23. organization in the design of computer menus: A two-experi- Kiger, J. I. 1984. The depth/breadth trade-off in the design of ment study. Journal of the American Society for Information menu-driven user interfaces. International Journal of Man- Science 44: 393–97. Machine Studies 20: 201–13. Crawford, W. 1987. Patron access: Issues for online catalogs. Kruk, R. S., and P. Muter. 1984. Reading of continuous text on Boston: G. K. Hall. video screens. Human Factors 26: 339–45. Crawford, W., L. Stovel, and K. Bales. 1986. Bibliographic displays Larson, R. R. 1991a. Between Scylla and Charybdis: Subject in the online catalog. White Plains, N.Y.: Knowledge Industry. searching in the online catalog. Advances in Librarianship 15: Cutter, C. A. 1904. Rules for a dictionary catalog. 4th ed. 175–236. Washington, D.C.: GPO. ———. 1991b. The decline of subject searching: Long term trends Davis, C. H., and D. Shaw. 1989. Comparison of retrieval system and patterns of index use in an online catalog. Journal of the interfaces using an objective measure of screen design effec- American Society for Information Science 42: 197–215. tiveness. Library and Information Science Research 11: Leazer, G. H. 1993. A conceptual plan for the description and con- 325–34. trol of bibliographic works. Ph. D. diss., Columbia University. Dervin, B., and M. Nilan. 1986. Information needs and uses. In Mandel, C. A. 1985. Enriching the library catalog record for sub- ARIST 21, edited by Martha Williams. White Plains, N.Y.: ject access. Library Resources & Technical Services 29: 5–15. Knowledge Industry, 2–33. Marchionini, G. 1992. Interfaces for end user information seeking. Dumas, J. S. 1988. Designing user interfaces for software. Journal of the American Society for Information Science 43: Englewood Cliffs, N.J.: Prentice Hall. 156–63. Fidel, R. 1984. Online searching styles: A case study-based model Marcus, A. 1982. Typographic design for interfaces of information of searching behavior. Journal of the American Society for systems. In Human factors in computer systems. New York: Information Science 35: 211–21. Association for Computing Machinery. Ford, N., and R. Ford. 1993. Towards a cognitive theory of infor- Marcus, A., W. Cowan, and W. Smith. 1989. Color in user interface mation processing: An empirical study. Information design: Functionality and aesthetics. ACM SIGCHI Bulletin: Processing & Management 29: 569–85. 25–27. Foss, C. L. 1989. Tools for reading and browsing hypertext. Markey, K. 1985. Subject-searching experiences and needs of Information Processing & Management 25: 407–18. online catalog users: Implications for library classification. Frei, H. P., and J. F. Jauslin. 1983. Graphical presentation of infor- Library Resources & Technical Services 29: 34–51. mation and services: A user-oriented interface. Information Martin, T. H. 1974. A feature analysis of interactive retrieval sys- Technology: Research and Development 2: 23–42. tems. Stanford, Calif.: Stanford University Institute for Galitz, W. O. 1989. Handbook of screen format design. 3d ed. Communication Research. Wellesley, Mass.: QED Information Sciences, Inc. Matthews, J. R. 1985. Suggested guidelines for screen layouts and Gregor, D., and C. A. Mandel. 1991. Cataloging must change! design of online catalogs. In Online catalog screen displays: A Library Journal 116 (April 1): 42. series of discussions. Washington, D.C.: Council on Library Hancock-Beaulieu, M., S. Robertson, and C. Neilson. 1991. Resources. Evaluation of online catalogues: Eliciting information from Millsap, L., and T. E. Ferl. 1993. Search patterns of remote users: the user. Information Processing & Management 27: 523–32. An analysis of OPAC transaction logs. Information Technology Harman, D. 1992. User-friendly systems instead of user-friendly & Libraries 12: 321–43. front-ends. Journal of the American Society for Information Morehead, D. R., and W. B. Rouse. 1982. Models of human behav- Science 43: 164–74. ior in information seeking tasks. Information Processing & Hensley, R. 1991. Learning style theory and learning transfer prin- Management 18: 193–205. ciples during reference interview instruction. Library Trends Noerr, P. L., and K. T. Bivins Noerr. 1985. Browse and navigate: An 39: 203–9. advance in database access methods. Information Processing Hildreth, C. R. 1982. Online public access catalogs: The user inter- & Management 21: 205–13. face. Dublin, Ohio: OCLC. Norman, D. A. 1990. The design of everyday things. New York: Hsieh-Yee, I. 1993. Effects of search experience and subject Doubleday. knowledge on the search tactics of novice and experienced Olsen, K. A., et al. 1993. Visualization of a document system: The searchers. Journal of the American Society for Information VIBE system. Information Processing & Management 29: Science 44: 161–74. 69–81. 46 Thomas LRTS 45(1)

Palmer, R. P. 1972. Computerizing the card catalog in the univer- online searching project, report number 2. Journal of the sity library: A survey of user requirements. Littleton, Colo.: American Society for Information Science 44: 273–91. Libraries Unlimited. Smiraglia, R. P. 1992. Authority control and the extent of derivative Prorak, D., T. Gottschalk, and M. Pollastro. 1994. Teaching relationships. Ph.D. diss., University of Chicago. method and psychological type in bibliographic instruction: Thomas, David H. 1997. The effect of interface design on item Effect on learning and confidence. RQ 33: 484–95. selection in an online catalog. Ph.D. diss., University of Rubin, J. 1994. Handbook of usability testing: How to plan, design, Pittsburgh. and conduct effective tests. New York: Wiley. Troll, D. A. 1995. What’s hot and what’s not. College & Research Seal, A. 1983. Experiments with full and short entry catalogues: A Libraries News 56: 236–39. study of library needs. Library Resources & Technical Trollip, S., and G. Sales. 1986. Readability of computer-generated Services 27: 144–55. fill-justified text. Human Factors 28: 159–63. Seymour, S. 1991. Online public access catalog user studies: A Tullis, T. S. 1981. An evaluation of alphanumeric, graphic, and review of research methodologies, March 1986–Nov. 1989. color information displays. Human Factors 23: 541–50. Library and Information Science Research 13: 89–102. ———. 1983. The formatting of alphanumeric displays: A review Shaw, D. 1991. The human-computer interface for information and analysis. Human Factors 25: 657–82. retrieval. In ARIST 26, edited by Martha Williams. Medford, ———. 1988. A system for evaluating screen formats: Research N.J.: Learned Information Inc., 155–95. and application. In Advances in human-computer interaction, Shires, N. L., and L. P. Olszak. 1992. What our screens should look vol. 2., edited by H. R. Hartson and D. Dix. Norwood, N.J.: like: An introduction to effective OPAC screens. RQ 31: Ablex: 214–86. 357–69. Yee, M. M. 1991. System design and cataloging meet the user: Shneiderman, B. 1992. Designing the user interface: Strategies for User interfaces to online public access catalogs. Journal of the effective human-computer interaction. 2d ed. Reading, Mass.: American Society for Information Science 42: 78–98. Addison-Wesley. Siegfried, S., M. J. Bates, and D. N. Wilde. 1993. A profile of end user searching behavior by humanities scholars: The Getty 45(1) LRTS 47

The Year’s Work in Cataloging, 1999

Amy K. Weiss and Timothy V. Carstens

The challenge of cataloging Web sites and electronic resources was the most important issue facing the cataloging world in the last year. This article reviews attempts to analyze and revise the cataloging code in view of the new electronic environment. The difficulties of applying traditional library cataloging standards to Web resources has led some to favor metadata as the best means of providing access to these materials. The appropriate education and training for library cataloging personnel remains crucial during this transitional period. Articles on user understanding of Library of Congress subject headings and on cataloging practice are also reviewed.

his review of the literature on cataloging for 1999 reveals a discipline in tran- Tsition. The “Year’s Work in Cataloging” articles that have appeared in earlier issues of Library Resources & Technical Services have been overviews of a relatively confined universe of print materials. Like them, this article surveys the many significant contributions to cataloging that have recently appeared in the print literature. It also, however, examines working documents publicly available on the Web that were created by the Association for Library Collections & Technical Services Committee on Cataloging: Description and Access (CC:DA), the Joint Steering Committee for the Revision of AACR (JSC), and by other bod- ies responsible for the creation of cataloging rules. The JSC is now in the process of making the most important revision to the cataloging code since the adoption of AACR2 in 1981. Without an examination of these materials it is not possible to fully grasp the scope of the major changes on the horizon in cataloging. The Internet, still a novelty in 1993 when the last “Year’s Work in Cataloging” articles were published, has become the single most important phenomenon in contemporary librarianship. The pressure and problems of integrating Web materials into the library have changed the environment in libraries enormously, and the full impact and implications of the Web cannot at this juncture be accounted. The mercurial, infinitely flexible Web forces us to examine in depth all aspects of current cataloging, from the difference between monographs and serials, to the relationship of monographic materials to one another, to the continued viability of the Machine Readable Cataloging (MARC) communications format, to the underpinnings of the Anglo-American Cataloging Rules themselves. Amy K. Weiss ([email protected]) is Principal Cataloger, Appalachian State University, Boone, North Carolina, Theory and Timothy V. Carstens (carstens@ wcuvax1.wcu.edu) is Head of Cata- loging and Acquisitions, Western In response to problems with the code that have become more apparent in the Carolina University, Cullowhee, North current electronic environment, a number of researchers have attempted to Carolina. examine the underlying assumptions and structure of the cataloging rules with a Manuscript received June 2, 2000; accepted for publication August 5, view toward establishing the foundations for “more logical and comprehensive 2000. cataloging codes” (Taniguchi 1999, 448). Taniguchi studies the structure of the 48 Weiss and Carstens LRTS 45(1)

rules using conceptual modeling. He analyzes a variety of Revising AACR cataloging rules, chiefly from AACR2, 1988 revision, chapter 1, in terms of whether the rules promote or “orient” A number of critical problems with the code, as it is cur- toward cost-effectiveness, identity, contents, or consistency. rently written, were illuminated at the conference. Two crit- Delsey’s work, “The Logical Structure of the Anglo- ical problems, in the terminology of the CC:DA, are known American Cataloging Rules,” is probably the most influen- as the multiple characteristics problem and the format vari- tial and comprehensive examination of the structure of the ation problem (ALCTS CC:DA 1999, 1). The multiple char- cataloging rules. Delsey stated, “the methodology used in acteristics problem results from the fact that, in addition to this study is derived from techniques used in system devel- physical format, every document potentially also has the fol- opment projects . . . entity-relationship and object-oriented lowing aspects: type of publication, published vs. unpub- models [which] are used as the basis for identifying the key lished, fundamental content, and method of management entities or objects about which an organization needs to and reproduction. AACR2 rule 0.24 appears to give prima- keep data and clarifying the data-related business rules that cy to the “physical carrier” aspect. In other words, a video- apply within the organization prior to designing the layout of cassette is cataloged using chapter 7 of the rules because it databases to support the organizations’ business activities” is a recording of visual images on magnetic tape. Delsey (Delsey 1998, 1). Šauperl and Saye analyzed previous (1998) has demonstrated that AACR2 is inconsistent about research on the possibility of creating expert systems for the physical carrier principle. For example, cartographic descriptive cataloging and summarize the obstacles to the information—whether on paper, transparencies, or three- creation of such a system. They concluded that inconsisten- dimensional models—is cataloged using chapter 3 because cies in the language and structure of the code are major dif- of the cartographic content of the items. In this case, the ficulties for the creation of expert systems, but consider that content of the item, not the carrier, determines its category. cataloger experience and the ability to interpret rules are In addition, AACR2 makes the unwarranted assumption also significant (Šauperl and Saye 1999). that each physical carrier is restricted to only one of the class Relationships in the catalog were the focus of Leazer of materials chapters in AACR2 (ALCTS CC:DA Task and Smiraglia’s work on the complexity of bibliographic rela- Force on Rule 0.24 [ALCTS CC:DA TF 0.24] 1999). This tionships. They found that many works are part of biblio- means that a relevant rule cannot be applied when it is not graphic families and have a system of relationships that are part of the class of material chapter being used to catalog the not well explored in the library catalog, since uniform titles document. Confusion is also caused by the fact that many and uniform name headings do not express all aspects of different types of content can often be put on the same car- interrelation (Leazer and Smiraglia 1999). Works that are rier. For example, maps, music, and text can all be digital. In canonical (or popular) have a particularly complex set of some cases, the scope notes in AACR2 chapters appear to relationships (Smiraglia and Leazer 1999). exclude altogether certain categories of materials from treat- Seymour Lubetzky remains a powerful influence on ment by AACR2 (ALCTS CC:DA TF 0.24 2000). those studying the form of the catalog and the form of cata- The format variation (or multiple versions) problem loging rules. Leazer mentioned him as an influence (St. relates to another function of rule 0.24. Rule 0.24 is the only Lifer 1999). Taniguchi also references him. His paper on place in AACR2 that addresses the question of whether to “form headings” (currently either uniform titles or corporate create a new record if a record for a similar item already exits headings), written for an ALA committee in the mid-1950s, in the catalog. As it is now written, the rule implies that a new is noteworthy since it deals with the problem of modifying record should be created whenever there is any variation in standards to meet a presumed difference in circumstance. “physical carrier” between two documents, even if the docu- Lubetzky’s analytical approach remains relevant, and his ref- ments contain the same intellectual or artistic content erences to the earliest history of cataloging serve to remind (ALCTS CC:DA TF 0.24 1999). For example, under the cur- us how much things stay the same even though the technol- rent rules, separate records are created for the film and video ogy of the catalog has changed. versions of a feature motion picture. Separate records are While not published in 1999, the pivotal recent works also created for the print, microfilm, and online versions of a on the nature of the code continue to be the papers pre- periodical, even if all three versions contain exactly the same sented at the International Conference on the Principles content. The growth in the number and variety of electronic and Future Development of AACR held in Toronto in 1997. resources has made this problem more acute since electron- The conference proceedings, published in 1998, include ic resources are capable of carrying so many different types important works by Hirons and Graham, Yee, Howarth, and of content. As a result of the format variation problem, the Delsey. These papers have become the philosophical under- catalog user may have to wade through many catalog records pinning for much of the work currently being done on the that represent essentially the same thing. In their local cata- cataloging codes. logs, many libraries are already adding holdings for various 45(1) LRTS The Year’s Work in Cataloging, 1999 49

versions of serials to only one catalog record. Even though (Hirons 2000). These new rules, if adopted, should allow this local practice has been sanctioned by the Cooperative AACR2 to handle more effectively this rapidly growing cat- Online Serials Program (CONSER), it is still contrary to egory of resources. AACR2 rules. Hirons’s report also recommended a series of modifica- Another problem that has become especially acute with tions to AACR2 to deal with the fact that continuing the advent of electronic resources resides in AACR2’s con- resources change continually over time (Hirons 1999). As cept of seriality. Because they do not have a predetermined noted above, the current rules only allow the cataloger to conclusion, Web sites and databases exhibit many of the consider a “snapshot” of the evolution of these resources. attributes of serials. Yet these resources are not issued in Hirons’s solution is to require catalogers to consider the successive parts and do not bear numeric or chronological whole publication, not just a single issue or manifestation (as designations, which are the other AACR2 attributes of seri- in the case of integrating resources). Proposed changes ality. Since these electronic resources lack the other “serial” emphasize identification of the whole publication over attributes, they cannot be treated as serials by AACR2; yet merely transcribing information from the most current man- because of their continuing nature they cannot be handled ifestation. Hirons recommended focusing on the description effectively as monographs. This means that there is now a of variable information, ignoring obvious errors in titles, large and growing category of bibliographic resources that minor title changes, and most “other title information” when cannot be handled effectively by the code. cataloging continuing resources. Hirons would require the A final problem with the code is that it assumes that the identification of all of the item(s) used as the basis of content of a document is permanently fixed within a physi- description (or, in the case of integrating resources, all times cal object. Again, this is not necessarily the case with that the resource was viewed). She would also apply to inte- remotely accessible electronic items, which can change from grating resources and the use of notes indicating other titles one viewing to another. Although the current rules allow the by which a resource is known. Her report recommended cataloger to handle this material by taking a “snapshot” that the source for the title or statement of responsibility (Delsey 1998, 35) of the item while at the same time leaving should be the latest issue or iteration of a resource. Closely certain variables vague, this technique is not adequate connected with this suggestion is the recommendation that because there is no way to know when content might have a uniform title be added to successively cataloged titles in changed or whether the “snapshots” compiled by various order to provide these resources with a stable title. These catalogers will be sufficiently similar to identify the item. two recommendations would provide the user with a more A number of different committees and organizations holistic approach to the whole work—i.e., both earliest and have been working on the above-mentioned problems with latest titles—while still providing stability. JSC has now the code. The definition of seriality was tackled in a report endorsed many of these changes. The use of title informa- prepared for the JSC by Hirons and others entitled tion from the latest piece in hand was endorsed for integrat- “Revising AACR2 to Accommodate Seriality” (1999). ing resources but not for traditional serials (JSC 1999). Hirons began by distinguishing “finite” publications from Another major change recommended by Hirons and intend- “continuing” publications. Finite publications are complete ed to emphasize the whole work is the elimination of the or are intended to be complete and include traditional concept of chief source of information for continuing monographs. Continuing publications are intended to be resources. JSC is currently exploring the possibility of mak- continued for an indeterminate period and include tradi- ing the entire resource the chief source of information for all tional serials. They also include a conceptually new type of types of materials, not just for continuing resources (JSC publication called integrating resources (though certain 1999). types of integrating resources can also be finite publica- CC:DA has recommended a two-pronged solution to tions). Hirons defined integrating issuance as “a form of the multiple version problem. issuance that describes bibliographic resources that are It suggested that AACR2 should be reorganized by added to or changed by means of updates that do not remain International Standard Bibliographic Description (ISBD) discrete and are integrated into the whole” (Hirons 1999). area so that all rules related to title, edition, etc., appear Integrating resources include loose-leaf publications, Web together (ALCTS CC:DA 1999, 4). The new arrangement sites, and databases. Hirons also recommended refinement would allow the cataloger to use every rule relevant to the of the AACR2 definition of serial to make the distinction resource being cataloged and does not restrict the cataloger between a traditional serial and an integrating resource to the rules in only one class of material chapter. It also rec- clearer. The JSC has endorsed the concept of integrating ommended that rule 0.24 be rewritten so that all aspects of resources and Hirons has been commissioned by the JSC to a document are brought out and so that physical carrier is no propose specific descriptive rules to deal with them (JSC longer given primacy. The key first sentence of the proposed 1999). The proposed revisions are now available for review rule revision reads, “It is important to bring out all aspects 50 Weiss and Carstens LRTS 45(1)

of the item being described, including its content, its carri- tronic resources, it is not surprising that the section of the er, its type of publication, its relationship to other expres- code that deals with these resources has also been undergo- sions of the same work, and whether it is published or ing a major revision. The process began at the international unpublished . . . all relevant aspects should be described level when the International Federation of Library with no one relevant aspect taking precedence over any Associations and Institutions (IFLA) published a major revi- other” (ALCTS CC:DA 1999, 5). JSC has endorsed a staged sion of the International Standard Bibliographic Description approach to the implementation of these revisions with for Electronic Resources [ISBD (ER), formerly ISBD some changes to rule 0.24 being made immediately and fur- (CF)]. CC:DA was charged with “harmonizing” AACR2 ther changes made should AACR2 be reorganized by ISBD chapter 9 with ISBD (ER). The CC:DA Task Force on area (JSC 1999). JSC is currently developing a prototype to Harmonization of ISBD (ER) and AACR2 has now pro- test the feasibility of reorganizing AACR2 by ISBD area. duced a final report, including recommendations for specif- JSC’s eventual disposition of the ISBD reorganization pro- ic changes to the code (ALCTS CC:DA Task Force on posal is highly uncertain at this time. Harmonization of ISBD (ER) with AACR2, 1999). These CC:DA has also made recommendations related to the recommendations parallel many of the ones made by those format variation problem. According to CC:DA, this prob- committees dealing with rule 0.24 and seriality and, indeed, lem results from confusion about the purpose of biblio- the chapter 9 recommendations influenced and were influ- graphic records. Should bibliographic records represent a enced by the other activities that have been under discus- particular expression of a work, or should they represent a sion. manifestation of a work? Expression is “the intellectual or The replacement of the term “computer file” with the artistic realization of a work in the form of alpha-numeric, term “electronic resource” throughout chapter 9 is the first musical, or choreographic notation, sound, image, object, change recommended by the CC:DA Task Force on movement, etc. or any combination of the above forms.” Harmonization. Also recommended were changes to the Expression excludes “aspects of physical form . . . that are definition of the edition of an electronic resource. Like the not integral to the intellectual or artistic realization of the CC:DA report on 0.24, the chapter 9 task force urged that a work as such” (ALCTS CC:DA TF 0.24 1999). resource be considered a new edition only if there are sig- Manifestation, on the other hand, is “the physical embodi- nificant changes in the intellectual or artistic content. ment of an expression of a work.” A manifestation repre- Statements such as new release, version, level, or update, as sents all the physical objects that bear the same well as changes in physical carrier, display format, or print- characteristics, in respect to both intellectual content and er-related file formats would require a new record only if physical form. When the production involves changes in there were also a change in content. Like Hirons, the task physical form, the resulting product is considered a new force recommended that the chief source of information for manifestation. Changes in physical form include changes an electronic resource be the entire resource, not the title affecting display characteristics . . . changes in physical screen. Other task force recommendations were concerned medium . . . changes in container” (ALCTS CC:DA TF 0.24 mainly with broadening the rules so as to make them more 1999). Currently, new records are created whenever there is applicable to the large variety of electronic resources that a new manifestation of a work. CC:DA has suggested that now exist. These recommendations included broadening the the rules be changed so that new records are created only if scope of the rules to include new types of electronic the expression changes (ALCTS CC:DA 1999). JSC has resources; creating more specific terms for use in the file reserved judgment on this recommendation (JSC 1999). characteristics area; and the inclusion of many more current Related to the format variation problem is the question examples, particularly those related to remote resources, in of whether minor variations in a resource require a new cat- the note fields. alog record. Although various entities (including OCLC and the LCRIs) have addressed this issue, AACR2 remains silent. JSC has endorsed this idea and has asked CC:DA to Summary of Proposed Changes draft an appendix to AACR2 defining what constitutes major and minor changes (JSC 1999). Major changes would The task forces and committees working on revision of require the creation of a new record; minor changes would AACR2 have proposed numerous fundamental changes to not. JSC has also asked the Library Association/British the cataloging code. Among the historic changes suggested Library to prepare a draft introduction to AACR2 that have been the defining of the new type of resource, the inte- would discuss the major/minor changes issue as well as other grating resource, and the creation of a new set of rules to general issues related to the entire code (JSC 1999). handle this type of resource. Another proposal was to create Since so many of the changes to the code discussed new records only when there is a new “expression” of a work above were stimulated by the increasing importance of elec- and not for every “manifestation.” A new appendix and 45(1) LRTS The Year’s Work in Cataloging, 1999 51

introduction to AACR2, as well as its reorganization by for Web authors to make their own metadata, and one ISBD area, are possible changes that would radically alter its Dublin Core enthusiast quoted in Chepsiuk implied that the appearance in addition to changing its content. Finally, the lack of rules for Dublin Core records makes them easy to many proposed changes to chapter 9 and to other parts of create (Chepesiuk 1999). Gradmann made the argument AACR2 to accommodate the slippery new forms of elec- that metadata is fundamentally different from cataloging, tronic resources would contribute to substantial reworking since it focuses on discovery of resources rather than of the code. Taken together, these rule revision efforts rep- description of resources, and further states that traditional resent some of the most important work being done in cat- description is not end user-oriented (Gradmann 1999). aloging today. Catalogers are urged to follow the progress of For some, these arguments are reductive of the prob- this revision process via the Web sites listed in the bibliog- lems of description and show nothing more than an incom- raphy of this article. plete grasp of both AACR2 and the capabilities of the MARC format. Gorman discussed the difference between framework standards such as MARC or metadata, which Metadata merely name elements in the description, and content standards, such as AACR2, which specify what content is includ- While the JSC and others work on revision of the cataloging ed in the catalog record (Gorman 1999). He shows the code and MARC to accommodate electronic resources, oth- strong similarity between MARC and the Dublin Core, and ers have tried new approaches to bibliographic control. warns against ignoring content standards or confusing con- These new approaches revolve around the nebulous concept tent with framework. Gradmann as well as Milstead and of metadata. Metadata is commonly defined as data about Feldman are guilty of this confusion, since they emphasize data. Aside from that rather glib definition, which is usually controlled language in metadata as essential for increased followed by the observation that even catalog cards are retrieval while failing to acknowledge that metadata stan- metadata, it has been rather difficult for noninitiates to dis- dards do not in any manner require controlled language cover what the furor over metadata is about and why meta- (Gradmann 1999; Milstead and Feldman 1999). Milstead data is important in information retrieval. Fietzer noted that and Feldman admitted that the variety of metadata stan- an ALCTS subcommittee has found “more than twenty dards limits the usefulness of metadata for searching. They working definitions in the course of its deliberations, each included the development of thesauri as one of the goals for one viable for the community that created it” (Fietzer 1999, unification (Milstead and Feldman 1999). 13). The confusion over how to define metadata has been Weinberg’s argument that the Web represents no more well analyzed by Hopkins, who stated, “Metadata is . . . than a wrinkle on traditional problems of indexing makes a information about, primarily, electronic resources, with that number of valuable points. She stated that the number of information being encoded according to some scheme Web sites is not as prohibitively large as has been portrayed, which is often electronic” (Hopkins 1999, 56). noting that there were at the time of her writing approxi- Compounding the confusion is the large number of metada- mately two million Web sites, compared to thirty-nine mil- ta standards. A number of articles attempt to explain meta- lion records for unique materials in OCLC, and thirty data. Ahronheim (1998) has written one of the most million in RLIN. She also compared the structure of Web comprehensive articles, and is a good place for the novice to sites to periodicals, noting that even a single edition of a start reading. It describes the markup languages used and a book can contain variants and that merely being mutable number of the most important standards and initiatives. (like a loose-leaf publication) does not make Web materials Summers provided links to current metadata projects and unusually novel. She exhorted librarians to work on cooper- gives instructions for viewing the metadata (Summers 1999). ative cataloging and indexing efforts, rather than spending The Dublin Core has become the emergent standard of time developing elaborate local Web pages that more often greatest interest to librarians because it is flexible, highly than not reference the same essential sites (Weinberg 1999). developed, and promoted by OCLC. Chepesiuk and Burk One feature of the current environment is a large num- both described the genesis of the Dublin Core and describe ber of local or specialized initiatives. Doran’s case study its key elements (Chepesiuk 1999; Burk 1999). described the development of metadata for a large corpo- Metadata is frequently offered as a workable alternative rate Intranet, using a metadata template and controlled lan- to full library cataloging for Web resources. Some argu- guage terms. Her case study is notable for the fact that it ments against using AACR2R and the MARC format for emphasizes how much work was involved in the retrospec- Web resources are noted in Chepesiuk (1999): MARC tive indexing of Web pages, and how difficult it was for staff, records are too difficult to make and it is not cost-effective who had volunteered to assist but who had no official time to make MARC records for often ephemeral Web materials. set aside for performing the work, to meet deadlines (Doran Gradmann (1999) noted that the Dublin Core was intended 1999). Veatch emphasized how traditional library tools such 52 Weiss and Carstens LRTS 45(1)

as collection development and MARC format can be uti- links with other relevant documents. Ellis and Vasconcelos lized in Web control. Veatch emphasized a selective explored how facet analysis might be used to bridge the gap approach: finding the resources that will augment the in retrieval capabilities between the word approach (used by library’s holdings in a given area and committing to full most search engines) and true subject searches (1999). MARC cataloging and any associated maintenance (Veatch Hopefully a concrete project can be created pursuing this 1999). Porter and Bayard described a similar project under- intriguing idea. taken at the University of Notre Dame. They emphasized The MARC format was developed in 1967, making it cataloging sites by “reliable” producers of information and ancient in automation terms. Hopkinson noted that the committing to using link-checking software as a database underpinnings of MARC, designed to facilitate the maintenance procedure. Few of their sites changed URLs exchange of information on digital tape, are dated. Despite or went under during the project. They noted that it took that drawback, the stability of MARC has created a near uni- catalogers approximately thirty minutes to catalog Web sites versal ability to exchange bibliographic information. initially and twenty minutes by the end of the project, which “MARC use has increased since people began to prophesy seems to invalidate the perception that current standards its death [in 1985]” (Hopkinson 1999, 18). Metadata, with its and processes are unworkable for Web materials (Porter and multiple standards and formats, may not prove as resilient as Bayard 1999, 391–92). MARC. One challenge of Internet cataloging, which appears to Metadata cannot currently be integrated into a standard be receiving little consideration, is that there is no definable library catalog. Instead, metadata must be attached to Web Internet catalog. From the various forms of the book cata- pages or it can be confined to a separate database of meta- log, through the card catalog, and even through contempo- data records. This limitation leads to the issue of how many rary automated systems, there is a known universe of places and in how many ways it is advisable to provide access possible searching behaviors and strategies limited by the to electronic resources. It is ironic that at a time when even nature of the catalog itself. The simplicity of the card cata- the largest research libraries have realized that a purchased log is the source of its nostalgic appeal. The Internet’s end- ILS is more efficient than attempting to maintain a propri- less sprawl can be searched by any number of tools whose etary system, that even many small libraries feel obliged to search strategies can be ambiguous. Because of the diverse create proprietary home pages to provide access to elec- algorithms of various search engines it is difficult to know tronic materials. In some cases library Web pages have where to locate and how to format metadata for optimum evolved into a shadow catalog—a nontransferable effort of user searching. Turner and Brackbill’s (1998) research on the sort decried by Weinberg. how the use of embedded metadata (specifically the HTML META tag) affects retrieval of Web pages is significant. Pages with keyword metatags were ranked as “more rele- Education of Catalogers and Cataloging Staff vant” to a search designed to retrieve the pages than pages without the keyword metatags. However, another type of The realization that library standards may be useful in metatag (a description metatag) exhibited no improvement organizing the chaotic Internet means that the skill of cata- in retrieval, apparently because the search engines stopped logers may be in demand “Now that Cataloging Is Cool!” as examining the document for multiple occurrences of the one article trumpets (Garman 1999, 6). This excitement has term after examining the description tag (Turner and not, unfortunately, translated into an interest in cataloging as Brackbill, 1998). Valauskas’s examples of search engine fail- an academic discipline. A study by Spillane showed that in ures in retrieving an electronic article containing imbedded 1998 only 52.1% of library schools required a cataloging metadata are also instructive. Drabenstott et al. (1999) cited course, compared to 77.1% in 1986. This discouraging sta- Bates as stating that successful catalog searchers are those tistic is somewhat offset by the fact that the number of cat- who understand the structure of the catalog rather than aloging courses taught overall has risen (Spillane 1999). The those who are experts on the topic that they are searching. removal of required cataloging courses would seem to por- If so, the diversity and complexity of Web search tools may tend that fewer prospective librarians will have exposure to continue to confound catalogers in their search for the best cataloging and therefore not only will fewer people become means of access through metadata of whatever type. catalogers, but fewer librarians overall will have a solid grasp Most authors, despite their vague longing for controlled of how to use the catalog. language, have focused on the descriptive aspects of meta- Bordeianu and Seiser’s article discussed educational data rather than subject retrieval. Interesting new ideas for requirements for paraprofessional catalogers in academic subject access include Dovey’s short article on meta-objects, libraries. There is no single educational standard for para- which proposes intelligent documents that read and under- professional catalogers, but most ARL libraries do not stand their own imbedded metadata and dynamically form require postsecondary education for copy catalogers, 45(1) LRTS The Year’s Work in Cataloging, 1999 53

although many accept postsecondary education in place of implementation. Hemmasi et al. give the history of subfield experience (Bordeianu and Seiser 1999). There is a general v implementation, and admit that the controversial v is only assumption of on-the-job training for professionals and the first step in a larger implementation of form headings paraprofessionals alike. Sill’s article offered a proposed “cur- that will have to include implementation of search capabili- riculum” for on-the-job training for a professional serials ties by the creators of integrated library systems (Hemmasi, cataloger (1999). Wiles-Young and Novak’s article expressed Miller, and Lasater 1999). concern for loss of training expertise, as professional librari- Harvard University has already implemented MARC ans in cataloging departments become managers (1999, field 655 in its catalog, in response to requests from refer- 450). ence librarians. Beall discussed the use of form/genre head- It is ironic that formal training in cataloging standards is ings in the Hollis catalog. The form/genre terms in the considered irrelevant to the presumed creators of the new catalog are derived from multiple thesauri and include many information infrastructure. In contrast, Banks convincingly unverified headings from copy records, activated when the showed how learning the rudiments of cataloging assists field was implemented for indexing and display. The use of users in locating needed materials in the catalog (1999). form/genre terms to limit searches is still largely confined to the reference users, but it appears that they do offer a useful way to improve access and search precision. Hopefully, Subject Cataloging other libraries will follow suit in adding 655 to their search indexes, which would generate impetus to create the stan- The most significant research of the year was the dard thesaurus called for by Beall (1999). Drabenstott study of user understanding of Library of Congress subject headings. Results of the survey, as published and analyzed in several journals, showed a uniformly Practice low level of understanding of LCSH, with a 31% accuracy rating for child users and a 39% accuracy rating for adult While the Internet monopolized much of recent cataloging users (Drabenstott, Simcox, and Fenton 1999b). Using a literature, other formats of material also received attention. similar methodology, it was shown that expert users (i.e., Freeborn (1999) discussed some unusual problems in 3-D librarians) were more accurate at assessing the meaning of audiovisual cataloging, demonstrating another case where LC subject headings, but that reference and technical serv- strict reading of AACR2 may fail to illuminate the frustrat- ices librarians still scored rather low rates of 52% and 55% ed cataloger. respectively (Drabenstott, Simcox, and Williams 1999). The two volumes of Cataloging and Classification Drabenstott et al. argue against eliminating LCSH, but Quarterly devoted to cataloging of cartographic materials point out that the subject headings they asked the partici- are an excellent addition to the complex area of map cata- pants to analyze were extremely complex, comprising three loging practice. Even with two volumes, the editors admit- or more concepts. It may be that fewer concepts should be ted to gaps in the coverage of topic (Andrew and Larsgaard worked into single headings. Another implication is that the 1999), which speaks to the need for a revision of the long order of headings does not seem to make much difference out-of-print Cartographic Materials (Stibbe 1982) and some to any group of users. Therefore a standard order of subdi- simplification in map cataloging practice. visions would save money and time in cataloging depart- Frohnsdorff’s article is that rare bird: an entertaining ments (Drabenstott, Simcox, and Fenton, 1999b). Certainly article on cataloging. It explores problems with the veracity given the relative flexibility of contemporary OPACS, with of publication information contained in erotica and bootleg their ability to invert headings or search keyword compo- recordings. Because their creators are usually operating nents, this suggestion appears sensible. barely within the law or outside of it, publication informa- Lundgren and Simpson surveyed graduate student per- tion on these materials is generally spurious. ceptions of the usefulness of various aspects of bibliograph- With the Library of Congress’s move to Pinyin roman- ic description for Web materials. While they regarded ization, issues of language transcription and display have subject terms as useful, the graduate students preferred been highlighted. Riedlmayer’s article contains dramatic summary notes or abstracts (Lundgren and Simpson 1999). examples of how the failure of online catalogs to display ver- One might conclude that summary notes are simply less nacular alphabets and diacritics affects retrieval and access nebulous than subject strings and do not require special to materials. His examples range from the familiar (the knowledge to interpret. many versions of the name Qaddafi) to the less familiar but The MARC format subject subfield was finally imple- possibly more egregious (showing how the removal of dia- mented in February 1999, but not without controversy, as critics from Hungarian words reduces the language to gib- reflected in the postings to AUTOCAT that followed the berish) (Riedlmayer 1999). Automated authority control 54 Weiss and Carstens LRTS 45(1)

processes strip diacritics from headings as part of the nor- Ayers, F. H. 1999. Time for a change: A new approach to catalogu- malization process for comparing headings. Bolick demon- ing concepts. Cataloging & Classification Quarterly 28, no. 2: strated how this can lead to erroneous name headings when 3–16. applied to Chinese-language materials (1999). Ayers, F. H., L. P. S. Nielsen, and M. J. Ridley. 1999. BOPAC 2: A There have been several articles this year on archival new concept in OPAC design and bibliographic control. Cataloging & Classification Quarterly 28, no. 2: 17–44. cataloging practice, most notably Hamburger’s useful case Baker, Nicholson. 1999. A couple of codicils about San Francisco. study of three grant-funded retrospective conversion proj- American Libraries 30, no. 3 (Mar.): 35. ects. Hamburger’s conclusions on administration, staffing, Banks, Julie. 1999. Teaching basic cataloging concepts to non- and record quality are worthwhile reading for anyone plan- library science students. Journal of Education for Library and ning a retrospective cataloging project (1999). Information Science 40, no. 2: 118–21. In summation, we stand on the threshold of a new era. Beall, Jeffrey. 1999. Indexing form and genre terms in a large aca- The momentum is in place. In the next few years, there will demic library OPAC: The Harvard experience. Cataloging & undoubtedly be sweeping changes made to our cataloging Classification Quarterly 28, no. 2: 65–71. rules, to our online catalogs and the records contained in Benaud, Claire Lise, Sever Bordeianu, and Mary Ellen Hanson. them, and to the structure of subject access and analysis. 1999. Cataloging productivity standards in academic libraries. The question of whether we will make things better for Technical Services Quarterly 16, no. 3: 43–67. Bolick, Hsi-Chu. 1999. Problems in the establishment of users is an open one. It is hoped that more research will be nonunique Chinese personal headings with special reference performed to determine how users categorize and locate to NACO guidelines and vendor-supplied authority control. information and materials and guide us in the change Library Resources & Technical Services 43: 95–105. process. Bordeianu, Sever, and Virginia Seiser. 1999. Paraprofessional catalogers in ARL libraries. College and Research Libraries 60, Bibliography no. 6: 532–40. Brisson, Roger. 1999. Online documentation in library technical Ahronheim, Judith R. 1998. Descriptive metadata: Emerging stan- services. Technical Services Quarterly 16, no. 3: 1–20. dards. Journal of Academic Librarianship 24, no. 5: 395–404. ———. 1999b. The world discovers cataloging: A conceptual ———. 1999. Technical services management issues in the meta- introduction to digital libraries. Journal of Internet Cataloging data environment. Technicalities 19, no. 6 (June–July): 4–6. 1, no. 4: 3–30. Andrew, Paige G. 1999. Cataloging the contemporary printed atlas. Burk, Alan. 1999. Metadata and the Dublin Core. Feliciter 45, no. Cataloging & Classification Quarterly 27, no. 1–2: 147–64. 1: 60–65. ———. 1999b. A survey technique for map collection retrospec- Burnett, Kathleen, Kwong Bor Ng, and Soyeon Park. 1999. A com- tive conversion projects. Cataloging & Classification parison of the two traditions of metadata development. Quarterly 27, no. 3–4: 405–12. Journal of the American Society for Information Science 50, Andrew, Paige G., and Mary Lynette Larsgaard. 1999. Maps and no. 13: 1209–17. related bibliographic materials: Cataloging, classification, and Burton, Paul F. 1999. Information professionals and the World bibliographic control. Part I, Introduction. Cataloging & Wide Web. Online and CD-ROM Review 23, no. 2: 103–4. Classification Quarterly 27, no. 1–2: xvii–xx. Byrum, John D. 1999. ISBD (ER): An overview. International Anyomi, Mary E. 1999. Outsourcing cataloging functions in South Cataloguing and Bibliographic Control 28, no. 3 (July–Sept.): Carolina public libraries. Bottom Line 12, no. 1: 29–33. 65–66. Armstrong, HelenJane, and Jimmie Lundgren. 1999. Cataloging Carter, Christina E., Sever Bordeianu, and Nancy Dennis. 1999. aerial photographs and other remote-sensing materials. The real world of integrating electronic resources into a Web Cataloging & Classification Quarterly 27, no. 1–2: 165–227. OPAC. Serials Librarian 36, no. 3–4: 455–60. Association for Library Collections & Technical Services. Com- Chepesiuk, Ron. 1999. Organizing the Internet: The “core” of the mittee on Cataloging: Description and Access. 1999. Over- challenge. American Libraries 30, no. 1 (Jan.): 60–63. view and Recommendations Concerning Revision of Rule Childress, Eric, Erik Jul, and Eric Miller. 1999. Un-building a mys- 0.24. Accessed May 2, 2000, www.ala.org/alcts/organization/ tery. Journal of Internet Cataloging 1, no. 4: 45–51. ccs/ccda/tf-024h.pdf. Cole, Jim. 1999. Cataloging of digitized texts. Cataloging & Association for Library Collections & Technical Services. Com- Classification Quarterly 28, no. 3: 45–54. mittee on Cataloging: Description and Access. Task Force on Condron, Lyn. 1999. New department head’s staff introduction sur- Harmonization of ISBD (ER) and AACR2. 1999. Final vey. Cataloging & Classification Quarterly 28, no. 3: 95–102. Report. Accessed May 3, 2000, www.ala.org/alcts/organiza- Cooknell, Jennifer. 1999. Crossing the line: The development of tion/ccs/ccda/tf-harm3.pdf. archival standards. Art Libraries Journal 24, no. 2: 26–30. Association for Library Collections and Technical Services. Cortez, Edwin M. 1999. Use of metadata vocabularies in data Committee on Cataloging: Description and Access. Task retrieval. Journal of the American Society for Information Force on Rule 0.24. 1999. Report to CC:DA 1999 Annual. Science 50, no. 13: 1218–23. Accessed May 3, 2000, www.ala.org/alcts/organization/ccs/ Crawford, Walt. 1999. The card catalog and other digital contro- ccda/tf-024g.html. versies. American Libraries 30, no. 1 (Jan.): 52–58. 45(1) LRTS The Year’s Work in Cataloging, 1999 55

Croghan, Anthony. 1999. The rules of the cataloguing game. Ficter, Darlene. 1999. Administrative and factual metadata for Catalogue & Index 132: 6–7. intranets: Issues and options. Online 23, no. 6 (Nov.–Dec.): Davis, Harry O., and James S. Chervinko. 1999. Map cataloging 88–90. and classification: The basic who, what, and where. Fietzer, William. 1999. Working our way through wonderland: Cataloging & Classification Quarterly 27, no. 1–2: 9–37. Technical services and the sociology of metadata. Delsey, Tom, et al. 1998. The logical structure of the Anglo- Technicalities 19, no. 6 (June–July): 1, 13–15. American Cataloguing Rules. Part I, Drafted for the Joint Fleck, Nancy W., and Margaret Rust. 1999. MicroMARC for inte- Steering Committee for Revision of AACR. Accessed May 2, grated formats. Library Hi Tech 16, no. 2: 37–44. 2000, www.nlc-bnc.ca/jsc/aacrint.pdf. Formson, J. W. 1999. The impact of information technology on the DeMeule, Marah, and Gwen Gregory. 1999. Reconcilable differ- cataloguing process at the University of Botswana Library. ences: Archivists, catalogers, and manuscript materials. African Journal of Library Archives and Information Science Technical Services Quarterly 16, no. 4: 35–43. 9, no. 1: 17–26. DeSilva, S. M., and A. N. Zainab. 1999. Identifying and categoriz- Franzmeier, Gunter. 1999. Union catalogs of serials in Germany ing published conference proceedings. Malaysian Journal of between 1945 and 1970. Serials Librarian 35, no. 4: 111–17. Library and Information Science 4, no. 1 (July): 41–59. Freeborn, Robert. 1999. Cataloging of the weird. MC Journal 6, Doran, Kelly. 1999. Metadata for a corporate intranet. Online 23, no. 2. Accessed April 3, 2000, http://wings.buffalo.edu/ no. 1 (Jan.–Feb.): 43–50. publications/mcjrnl/v6n2/freeborn.html. Dorman, David. 1999. Metadata musings. American Libraries 30, Frohnsdorff, Gregory. 1999. Facts? Of publication: Cataloging no. 1 (Jan.): 102. problems posed by deceptive information. Library Resources Dorner, Dan. 1999. Cataloging in the 21st century. Part 1, & Technical Services 43: 213–22. Contextual issues. Library Collections, Acquisitions, and Garcha, Rajinder, and Lois Buttlar. 1999. Changing roles of cata- Technical Services 23, no. 4: 393–99. loguers in British academic libraries. Library Review 48, no. Dovey, Matthew. 1999. Meta-objects: An object-oriented approach 2: 66–72. to metadata. Ariadne 19 (Mar.). Accessed April 4, 2000, Garman, Nancy. 1999. Now that cataloging is cool. Online 23 http://ariadne.ac.uk/issue19/meta-objects. (Sept./Oct.), no. 5: 6. Drabenstott, Karen M., et al. 1999. Interpreting the findings of A Gilyard, Burt. 1999. Sandy Berman’s last stand. City Pages 20, no. study of library users and their understanding of subject head- 971 (July 14): 12–21. ings. Technicalities 19, no. 4 (Apr.): 13–15. Gorman, Michael. 1999. Metadata or cataloguing? A false choice. Drabenstott, Karen M., Bonnie Roeber Dede, and Melanie Leavitt. Journal of Internet Cataloging 2, no. 1: 5–22. 1999. The changes of meaning in subdivided subject headings. Gradmann, Stefan. 1999. Cataloguing vs. metadata: Old wine in Cataloging & Classification Quarterly 28, no. 3: 19–43. new bottles? International Cataloguing and Bibliographic Drabenstott, Karen M., Schelle Simcox, and Eileen G. Fenton. Control 28, no. 4 (Oct.–Dec.): 88–90. 1999. Do patrons understand Library of Congress Subject Hakala, Juha. 1999. Internet metadata and library cataloging. Headings? Technicalities 19, no. 1 (Jan.): 1, 7–11. International Cataloguing and Bibliographic Control 28, no. ———. 1999b. End user understanding of subject headings in 1 (Jan.–Mar.): 21–25. library catalogs. Library Resources & Technical Services 43: Hamburger, Susan. 1999. Life with grant: Administering manu- 140–60. scripts cataloging grant projects. The American Archivist 62: Drabenstott, Karen M., Schelle Simcox, and Marie Williams. 1999. 130–52. Do librarians understand the subject headings in library cata- Heaney, Mike. 1999. An interview with Tom Delsey. Cataloging & logs? Reference and User Services Quarterly 38, no. 4: 369–87. Classification Quarterly 28, no. 3: 3–18. Dunsire, Gordon. 1999. Bringing it all back home: Retrieval and Hearn, Stephen S. 1999. Metadata structures and authority con- access for the global information society. International trol. Technicalities 19, no. 6 (June–July): 7–9. Cataloguing and Bibliographic Control 28, no. 1 (Jan.–Mar.): Hemmasi, Harriette, David Miller, and Mary Charles Lasater. 13–14. 1999. Access to form data in online catalogs. ALCTS Online Ede, Stuart. 1999. Digital indigestion: Is there a cure? International Newsletter 10, no. 4 (July). Accessed April 3, 2000, Cataloguing and Bibliographic Control 28, no. 1 (Jan.–Mar.): www.ala.org/alcts/alcts_news/v10n4/formdat2.html. 3–7. Henshaw, Robin. 1999. The First Monday metadata project. Libri Edwards, Jennifer, and Amanda Xu. 1999. Eeee! Serials! Providing 49: 125–31. access to online serials. Serials Librarian 36, no. 3–4: 467–73. Hill, Linda L., et al. 1999. Collection metadata solutions for digital Ellis, David, and Ana Vasconcelos. 1999. Ranganathan and the library applications. Journal of the American Society for Net: Using facet analysis to search and organize the World Information Science 50, no. 13: 1169–81. Wide Web. ASLIB Proceedings 51, no. 1: 3–10. Hiller, Eileen. 1999. Scatter and save! An evaluation of the adop- Ercegovac, Zorana. 1999. Special topic issue: Integrating multiple tion of centralized classification and outsourcing. Against the overlapping metadata standards. Journal of the American Grain 11, no. 2: 18, 20–22. Society for Information Science 50, no. 13: 1165–68. Hirons, Jean, et al. 2000. Revising AACR2 to accommodate serial- Ferrari, Roberto C. 1999. The art of classification: Alternate classi- ity, rule revision proposals submitted to the Joint Steering fication systems in art libraries. Cataloging & Classification Committee for Revision of AACR. Accessed May 3, 2000, Quarterly 28, no. 2: 73–98. www.nlc-bnc.ca/jsc/ch12.pdf. 56 Weiss and Carstens LRTS 45(1)

———. 1999. Revising AACR2 to accommodate seriality, report to Lahary, Dominique. 1999. UNIMARC as a cataloguing format in the Joint Steering Committee for Revision of AACR. France. International Cataloguing and Bibliographic Control Accessed May 3, 2000, www.nlc-bnc.ca/jsc/ser-rep0.html. 28, no. 2: 50–52. Hirons, Jean, and Regina Reynolds. 1998. Proposal to adopt a mod- Landland, Kristen, ed. 1999. Ongoing entities: The impact of modified model C. Accessed April 11, 2000, www.lcweb.loc.gov/ ified model C on cataloging description. Serials Review 25, acq/conser/ModelC.html. no. 1: 95–108. ———. 1999. Seriality: It’s not just for serials any more. Library Lange, Holley R. 1999. Cataloging maps: Getting started. Review 48, no. 4: 163–68. Colorado Libraries 25, no. 1: 40–41. Holm, Liv Aasa. 1999. Authority control in an international context Larsgaard, Mary Lynette. 1999. Cataloging cartographic materials in the new environment. International Cataloguing and on CD-ROMs. Cataloging & Classification Quarterly 27, no. Bibliographic Control 28, no. 1 (Jan.–Mar.): 11–13. 3–4: 363–74. Holt, Brian. 1999. Presentation of UNIMARC on the Web: New Leazer, Gregory H., and Richard P. Smiraglia. 1999. Bibliographic fields including the one for electronic resources. International families in the library catalog: A qualitative analysis and Cataloguing and Bibliographic Control 28, no. 2 (Apr.–June): grounded theory. Library Resources & Technical Services 43: 47–48. 191–212. Hopkins, Judith. 1999. USMARC as a metadata shell. Journal of Lichtenbergova, Edita, and Bohdana Stoklasova. 1999. UNI- Internet Cataloging 2, no. 1: 55–68. MARC in Czech libraries. International Cataloguing and Hopkinson, Alan. 1999. Traditional communication formats: Bibliographic Control 28, no. 2 (Apr.–June): 52–54. MARC is far from dead. International Cataloguing and Liu, Jessica. 1999. Review and prospect for centralized cataloging in Bibliographic Control 28, no. 1 (Jan.–Mar.): 17–21. China. Cataloging & Classification Quarterly 28, no. 2: 57–64. Howarth, Lynne. 1999. Audio-visual cataloging: From orphan child Lubetzky, Seymour. 1999. On the use of form headings in an alpha- to Cinderella. Technicalities 19, no. 2 (Feb.): 1, 3–5. betical catalog. Library Quarterly 69, no. 2: 222–36. Hu, Jiajian. 1999. Chinese romanization in Library of Congress Lundgren, Jimmie, and Betsy Simpson. 1999. Looking through cataloging. Illinois Libraries 81, no. 4 (fall): 250–51. users’ eyes: What do graduate students need to know about Intner, Sheila. 1999. Back to the future of AV. Technicalities 19, no. Internet resources via the library catalog? Journal of Internet 2 (Feb.): 6–9. Cataloging 1, no. 4: 31–44. Irwin, Barbara. 1999. A fundamental of cataloguing efficiency: The Luttman, Stephen F. 1999. Good enough for jazz, or, successful automated batch method. Feliciter 45, no. 1: 14–15. music cataloging for non-musicians. Colorado Libraries 25, ISBD (CF) Review Group. 1997. ISBD (ER): International no. 2: 48–49. Standard Bibliographic Description for electronic resources; MacEwan, Andrew. 1999. Working with LCSH: The cost of cooper- Revised from the ISBD (CF) International Standard ation and the achievement of access. International Cataloguing Bibliographic Description for Computer Files. Munchen: K. and Bibliographic Control 28, no. 4 (Oct.–Dec.): 94–97. G. Saur. Madison, Olivia M. A. 1999. Standards in light of new technologies Jantz, Ronald C. 1999. An approach to managing vocabulary for functional requirements for bibliographic records. Inter- databases on the Web. Cataloging & Classification Quarterly national Cataloguing and Bibliographic Control 28, no. 1 28, no. 3: 55–66. (Jan./Mar.): 7–10. Joint Steering Committee for Revision of Anglo-American Manning, Ralph W. 1999. The Anglo-American Cataloging Rules Cataloguing Rules. 1999. Outcomes of meeting held in and their future. International Cataloguing and Bibliographic Brisbane, Australia, October 18–20, 1999. Accessed May 3, Control 28, no. 3 (July–Sept.): 68–71. 2000, www.nlc-bnc.ca/jsc/brisbane.htm. Marcus, Sara. 1999. How to catalog a human, or, these things must Kandoian, Nancy A. 1999. Cataloging early printed maps. be done delicately. American Libraries 30, no. 1 (Jan.): 94–95. Cataloging & Classification Quarterly 27, no. 3–4: 229–64. Martin, Rebecca. 1999. Notes for special collections in the MARC Kgosiemang, Rose Tiny. 1999. Retrospective conversion: The expe- format for holdings data: Subfield Z. Technicalities 19, no. 7 rience at the University of Botswana Library. Cataloging & (Aug.): 1, 8. Classification Quarterly 28, no. 3: 67–94. Martínez-Arellano, Filiberto Felipe. 1999. Subject searching in Kniffel, Leonard. 1999. Cataloger demands reprimand be rescind- online catalogs including Spanish and English material. ed. American Libraries 30, no. 4 (Apr.): 20–22. Cataloging & Classification Quarterly 28, no. 2: 45–46. Konovalov, Yuri. 1999. Cataloging as a customer service: Applying McDermott, Irene E. 1999. Running rings around the Web. knowledge to technology tools. Information Outlook 3, no. 9 Searcher 7, no. 4 (Apr.): 67–70. (Sept.): 25–27. McDonnell, Janice P., Wallace C. Koehler Jr., and Bonnie C. Kuhn, Tamara J. 1999. Classifying newspapers using Dewey Carroll. 1999. Cataloging challenges in an area studies virtual Decimal Classification. Library Resources & Technical library catalog (ASVLC): Results of a case study. Journal of Services 43: 106–13. Internet Cataloging 2, no. 2: 15–41. Kulczak, Deborah E. 1999. Name authority work for OCLC copy McEathron, Scott. 1999. The cataloging of globes. Cataloging & cataloging: Is it worth the effort? Cataloging & Classification Classification Quarterly 27, no. 1–2: 103–12. Quarterly 28, no. 1: 69–81. McIlwaine, I. C., and N. J. Williamson. 1999. International trends Kwansnik, Barbara H. 1999. The role of classification in knowledge in subject analysis research. Knowledge Organization 26, no. representation and discovery. Library Trends 48, no. 4: 22–47. 1: 23–29. 45(1) LRTS The Year’s Work in Cataloging, 1999 57

McKiernan, Gerry. 1999. Points of view: Conventional and “neo- Robertson, Michelle Martin. 1999. Customizing the organization of conventional” access and navigation in digital collections. local planning documents. Technicalities 19, no. 7 (Aug.): 6–7. Journal of Internet Cataloging 2, no. 1: 23–41. Rockwell, Ken. 1999. Problem areas in descriptive cataloging of sheet Mead-Donaldson, Robert. 1999. Classify your media! Technicalities maps. Cataloging & Classification Quarterly 27, no. 1–2: 39–63. 19, no. 1 (Jan.): 3–4. Roe, Sandy. 1999. Online subject access. Journal of Internet Medeiros, Norm. 1999. Making room for MARC in a Dublin Core Cataloging 2, no. 1: 69–78. world. Online 23, no. 6 (Nov./Dec.): 57–60. Rogers, Michael, and Norman Oder. 1999. Citing “deception,” Miller, Eric, Eric Childress, and Erik Jul. 1999. Making progress: Berman resigns from Hennepin County. Library Journal 124, The resource description framework. Journal of Internet no. 9 (May 15): 14. Cataloging 1, no. 4: 53–58. ———. 1999b. Hennepin County rebukes Berman. Library Milstead, Jessica, and Susan Feldman. 1999. Metadata: Cataloging Journal 124, no. 6 (Apr. 1): 13–16. by any other name. Online 23, no. 1 (Jan./Feb.): 25–31. Romero, Lisa, and Nancy Romero. 1999. Cataloging early atlases: ———. 1999b. Metadata: Projects and standards. Online 23, no. 1 A reference source. Cataloging & Classification Quarterly 27, (Jan./Feb.): 32–40. no. 3–4: 265–84. Molyneux, Robert. 1999. May you live in interesting times. Sandberg-Fox, Ann. 1999. Principal changes in the ISBD (ER). Technicalities 19, no. 2 (Feb.): 1, 10–13. International Cataloguing and Bibliographic Control 28, no. Moore, Susan. 1999. Navigating the G schedule. Cataloging & 3 (July–Sept.): 67. Classification Quarterly 27, no. 3–4: 375–84. Šauperl, Alenka, and Jerry D. Saye. 1999. Pebbles of the mosaic of Morris, Dilys E., and Gregory Wool. 1999. Cataloging: Librarian- cataloging expertise: What do problems in expert systems for ship’s best bargain. Library Journal 142, no. 11 (July 15): 44–46. cataloging reveal about cataloging expertise? Library Murtomaa, Eeva. 1999. The impact of the functional requirements Resources & Technical Services 43: 78–94. for bibliographic records recommendations on the ISBD (ER). Scheschy, Virginia. 1999. Outsourcing: A strategic partnership. Cataloging & Classification Quarterly 28, no. 1: 33–41. Technical Services Quarterly 16, no. 3: 31–41. ———. 1999b. The Net is revolutionizing cataloguing. Inter- Seymour, Chaim. 1999. Retrospective conversion at Bar-Ilan national Cataloguing and Bibliographic Control 28, no. 4 University. Technicalities 19, no. 7 (Aug.): 4–5. (Oct.–Dec.): 97–99. Sherwood, Arlyn. 1999. Enhancing in OCLC’s map format: A par- Nafito, Michael. 1999. The indexed Web: Engineering tools for ticipant’s view. Cataloging & Classification Quarterly 27, no. cataloging, storing, and delivering Web-based documents. 3–4: 429–41. Information Outlook 3, no. 2 (Feb.): 18–22. Sill, Laura A. 1999. Training for the new serialist. Library Parker, Velma. 1999. Cataloguing maps series and serials. Resources & Technical Services 43: 247–56. Cataloging & Classification Quarterly 27, no. 1–2: 65–101. Sitas, Anestis. 1999. Greek folk literature, poetry, folk songs, and ———. 1999b. MARC tags for cataloging cartographic materials. the Library of Congress PA (supplement) schedule. Cataloging & Classification Quarterly 27, no. 1–2: 5–8. Cataloging & Classification Quarterly 28, no. 1: 53–68. Porter, G. Margaret, and Laura Bayard. 1999. Including Web sites Smiraglia, Richard P., and Gregory H. Leazer. 1999. Derivative in the online catalog: Implications for cataloging, collection bibliographic relationships: The work relationship in a global development and access. The Journal of Academic bibliographic database. Journal of the American Society for Librarianship 25, no. 5: 390–94. Information Science 50, no. 6: 493–504. Prescott, Dorothy. 1999. Early maps with or in printed publica- Smits, Jan. 1999. Metadata: An introduction. Cataloging & tions: Description and access. Cataloging & Classification Classification Quarterly 27, no. 3–4: 303–19. Quarterly 27, no. 3–4: 285–301. ———. 1999b. Spatial metadata: An international survey on clear- Rahman, Salimah Abd, and Adibah Ahmad Rafae. 1999. inghouses and infrastructures. Cataloging & Classification Cataloging and treatment of materials written in Jawi: Quarterly 27, no. 3–4: 321–42. Experiences of Malaysia. International Cataloguing and Spillane, Jodi Lynn. 1999. Comparison of required introductory Bibliographic Control 28, no. 3 (July–Sept.): 73–75. cataloging courses, 1986 to 1999. Library Resources & Ralston, Rick, and Margaret A. Rioux. 1999. With feet planted Technical Services 43: 223–30. firmly in mid-air: Staff training for automation system migra- St. Lifer, Evan. 1999. Gregory Leazer: The new face of cataloging. tion. Serials Librarian 36, no. 3–4: 407–13. Library Journal 124, no. 8 (May 1): 42–44. Randall, Barbara Nichols. 1999. Spelling errors in the database: Stibbe, Hugo L. P. 1999. Cataloguing cartographic materials in Shadow or substance? Library Resources & Technical archives. Cataloging & Classification Quarterly 27, no. 3–4: Services 43: 161–69. 443–63. Reynolds, Regina Romano. 1999. Harmonizing bibliographic con- Stibbe, Hugo L. P., et al. 1982. Cartographic materials: A manual trol of serials in the digital age. Cataloging & Classification of interpretation for AACR2. Chicago: ALA. Quarterly 28, no. 1: 3–19. Studwell, William. 1999. Does the information explosion mean cat- Riedlmayer, András J. 1999. “Ghost cataloging” and other tales of aloging implosion? A re-exploration. Technicalities 19, no. 3 degraded access. Art Documentation 18, no. 1: 29–34. (Mar.): 13–14. Riesthuis, G. J. A., and R. Storm. 1999. GOO: The Dutch national ———. 1999b. Universal subject environment: Aspirations for a system for subject indexing. International Cataloguing and multinational, multicultural, and multilingual subject access Bibliographic Control 28, no. 4 (Oct.–Dec.): 91–93. system. Technical Services Quarterly 16, no. 3: 21–29. 58 Weiss and Carstens LRTS 45(1)

Su, Julie Tao. 1999. Keeping serial bibliographical records up-to- Weihs, Jean. 1999. Cardinal rule change and the OPAC. date in local online catalogs: A study on record change and Technicalities 19, no. 4 (Apr.): 1, 11–13. database maintenance. Serials Librarian 35, no. 4: 17–28. Weihs, Jean, ed. 1998. The principles and future of AACR: Summers, Ed. 1999. Chaos—show me the metadata! Against the Proceedings of the International Conference of the Principles Grain 11, no. 1: 79–82. and Future of AACR. Toronto, Ontario, Canada, October Sutton, Stuart A. 1999. Conceptual design and deployment of a 23–25, 1997. Ottawa: Canadian Library Association. metadata framework for educational resources on the Weimer, Katherine H. 1999. Subject analysis for cartographic Internet. Journal of the American Society for Information materials. Cataloging & Classification Quarterly 27, no. 3–4: Science 50, no. 13: 1182–92. 385–404. Swan-Hill, Janet. 1999. You may already know the answer. Serials Weinberg, Bella Hass. 1999. Improved Internet access: Guidance Librarian 36, no. 1–2: 225–45. from research on indexing and classification. Bulletin of the Taniguchi, Shoichi. 1999. An analysis of orientedness in cataloging American Society for Information Science 25, no. 2 rules. Journal of the American Society for Information Science (Dec.–Jan.): 26–29. 50, no. 5: 448–60. Weintraub, Jennifer, and Janet McKinney. 1999. The development Taylor, Arlene G. 1999. Where does AACR2 fall short for Internet and use of a genre statement for electronic journals. Serials resources? Journal of Internet Cataloging 2, no. 2: 43–52. Librarian 36, no. 3–4: 429–39. Thiry, Christopher J. J. 1999. Cataloging geologic sections. Welch, Grace D., and Frank Williams. 1999. Cataloguing digital Cataloging & Classification Quarterly 27, no. 1–2: 113–45. materials. Cataloging & Classification Quarterly 27, no. 3–4: Tsvetkova, Irina, and Vladimir Skvortsov. 1999. Adaptation of UNI- 343–62. MARC as Russian Exchange Format. International catalogu- Wendler, Robin. 1999. Branching out: Cataloging skills and func- ing and bibliographic control 28, no. 2 (Apr.–June): 54–56. tions in the digital age. Journal of Internet Cataloging 2, no. 1: Turner, Thomas P., and Lise Brackbill. 1998. Rising to the top: 43–54. Evaluating the use of HTML META tags to improve the Wiles-Young, Sharon, and Linda Novak. 1999. What happened to retrieval of World Wide Web documents through Internet the serials cataloger: Copy cataloging of serials. Serials search engines. Library Resources & Technical Services 42: Librarian 36, no. 3–4: 401–5. 258–71. Witt, Maria. 1999. Cataloguing CD-ROMs using the ISBD (ER) Valauskas, Edward J. 1999. Digital information and the Internet: rules—Examples of a French public library. Cataloging & The special case of electronic journals. International Classification Quarterly 28, no. 1: 21–31. Cataloguing and Bibliographic Control 28, no. 1 (Jan.–Mar.): Womble, Kathryn. 1999. Retrospective conversion and cataloging 14–17. of a major academic map collection: The University of Veatch, James R. 1999. Insourcing the Web. American Libraries Washington story. Cataloging & Classification Quarterly 27, 30, no. 1 (Jan.): 64–67. no. 3–4: 413–28. Verrill Smith, Rebecca. 1999. Dewey Decimal classification of folk- Yee, Martha M. 1999. One catalog or no catalog? ALCTS Online lore for children. Technicalities 19, no. 8 (Sept.): 4–6. Newsletter 10, no. 4 (July). Accessed Apr. 3, 2000, Wang, Daphne Hsu-Kuang. 1999. Collections without collective www.ala.org/alcts/alcts_news/v10n6/gateway_pap14.html. title. Journal of East Asian Libraries 117 (Feb.): 19–97. ———. 1999b. Guidelines for OPAC displays. ALCTS Online Warner, Amy J. 1999. A reaction to the findings of A study of Newsletter 10, no. 6 (Dec.). Accessed Apr. 3, 2000, library users and their understanding of subject headings. www.ala.org/alcts/alcts_news/v10n6/gateway_pap14.html. Technicalities 19, no. 4 (Apr.): 4–5. Zeng, Marcia Lei. 1999. Metadata elements for object description Weber, Mary Beth. 1999. Factors to be considered in the selection and representation: A case report from a digitized historical and cataloging of Internet resources. Library Hi Tech 17, no. fashion collection project. Journal of the American Society for 3: 298–303. Information Science 50, no. 13: 1193–208. Weibel, Stuart. 1999. The state of the Dublin Core metadata ini- Zuidema, Karen Huwald. 1999. Reengineering technical services tiative. Bulletin of the American Society for Information processes. Library Resources & Technical Services 43: 37–52. Science 25, no. 5 (June–July): 18–22. 45(1) LRTS 59

Book Reviews Margaret Rohdy, Editor

From Carnegie to Internet2: vice president of the Canadian firm lenges of serials that are still mostly Forging the Serials Future: Micromedia. He enumerates ten printed on acidic paper, decades after Proceedings of the North trends in the new information econo- preservation awareness became com- American Serials Interest my and ten strategies for success. This mon in libraries. Group, Inc.: 14th Annual might seem outside the traditional Between the plenary sessions and Conference, June 10–13, 1999. scope of serials work, but it reflects workshops are a series of “issues ses- Ed. P. Michelle Fiander, Joseph how the work of serialists has expand- sions.” These are case studies and C. Harmon, and Jonathan David ed in the last ten years. His thoughts broader discussions of current topics Makepeace. New York: Haworth make interesting reading even for that occupy a useful middle ground Information Pr., 2000. 398p. librarians not involved with the serials between the theoretical plenaries and $69.95 (ISBN 0-7890-1007-0); chain. the practicality of the workshops. Most $39.95 paper, (ISBN 0-7890- Balancing the visionary talks are of the 1999 sessions focused on elec- 1035-6). LC00-23223. NASIG’s many practical, serials- tronic issues. Several publishers The North American Serials Interest focused workshops that mine the described projects they had undertak- Group (NASIG) is unusual among broad and varied talents of NASIG’s en and what they had learned from library organizations because it is com- membership. While focused on local them. A diverse group of presenters pletely independent, unaffiliated with examples, NASIG workshops tend to tackled licensing issues from each any other body. It is also catholic in its transcend the “how we done it good” party’s perspective. There is an membership, embracing anyone inter- genre and provide insights that can be intriguing set of talks on advertising in ested in the serials world—librarians, more broadly applied. The workshop scholarly journals, with a communica- publishers, subscription agents, sys- topics show how NASIG’s focus has tions professor arguing in favor and a tem vendors. There are no institution- changed with the times. Many of the giant scientific journal publisher al or corporate members, so all players workshops present practical advice for sounding notes of caution. are equal. The result has been a handling the rise of electronic serials A few sessions highlighted plan- dynamic, productive organization that and databases. Electronic acquisitions ning and management skills for has gained respect in all corners of the methods were one major focus: there librarians. Mary Devlin, a profession- industry. Its annual conferences have were workshops on bundling and al communication specialist, present- become known as good places for aggregation, consortial licensing, push ed an issue session on nonverbal frank and interesting discussions, both technology, and desktop article deliv- communication in the work environ- formal and informal, of serials-related ery. At a licensing workshop, partici- ment. Two librarians from Maryland issues. Haworth Press has published pants found real examples to critique conducted a preconference on sce- the NASIG proceedings since their and a discussion of the major pitfalls of nario building as an alternative to beginning, both in The Serials licenses. strategic planning for libraries. Librarian and, as is its standard prac- On the other hand, some topics Library directors Glorianna St. Clair tice, as monograph separates. defy the passage of time. A popular of Carnegie Mellon University and The NASIG conference mixes workshop featured a panel discussion Rush Miller of the University of plenary speeches on general subjects, by two subscription agents, a publish- Pittsburgh spoke of the ACRL given by industry leaders, with practi- er, and a librarian on classic serials Redefining Scholarship Project, an cal workshops conducted by NASIG acquisitions questions: direct vs. attempt to describe and validate the members. In recent conferences, the agency orders, handling unsolicited research done by academic librarians. plenary sessions have become more material, switching from print to elec- The interest in these sessions also general and visionary, less focused on tronic versions. Most of the questions underscores the wide-ranging and serials and more on the new informa- might have been asked five years earli- far-sighted interests of serialists. tion society. This trend continues in er, and some even at the first NASIG The publication lag for any pro- the 1999 conference, with a particular- conference in 1986. Another work- ceedings volume can date material ly good essay by Stephen Abram, a shop focused on the preservation chal- quickly. NASIG’s material is not 60 Book Reviews LRTS 45(1)

immune; in fact, it can illustrate how the fray, aiming to assess the larger libraries as both document collections quickly some things change. Jean picture into which all the large and (using “document” generically) and Hirons of the Library of Congress and small individual digital initiatives systems in which users interact with Les Hawkins of the National Serials might fit. From Gutenberg to the the documents, much like libraries are Data Program presented one of the Global Information Infrastructure both discrete collections of materials first discussions of the proposed pulls together what we know now and systems in which users interact changes in the definition of “seriality” about the world of networked infor- with those materials; chapter 3 covers at NASIG in 1998. The proposal mation systems with what research familiar territory about methods of moved forward quickly and is now well and intuition lead us to anticipate in providing access to information, known and accepted in the serials the future. The author attempts to including the latest metadata schemas; community. The 1999 NASIG work- integrate what we know about tech- and chapter 4, on electronic publish- shop, however, still presents a succinct nology, human behavior, and policy ing, technology, and institutions, and clear summary of the proposal. regarding access to information. examines the life cycle of information Similarly, the preconference on meta- Borgman begins with definitions. and how developing technologies and data might seem less relevant now that First, she defines digital libraries as society’s institutions affect it. there have been many articles and both “a set of electronic resources and Chapters 5 and 6, “Why Are larger workshops on the topic, but the associated technical capabilities for Digital Libraries Hard to Use?” and NASIG report is still a good, straight- creating, searching, and using infor- “Making Digital Libraries Easier to forward introduction to metadata. mation” and an entity “constructed— Use,” contain the central core of ideas The special feature of the annual collected and organized—by [and for] for which the first four chapters pre- NASIG proceedings, as opposed to a community of users” (42). Then, she pare the reader. Here Borgman lays the rest of the serial literature, is its seeks to enlarge the reader’s vision still out her principal theses. Digital careful blend of theory and practice. further as well as to match it with the libraries were not easy to use in the NASIG’s program planning committee complexity of the real world, stating, “I past; but, now, design must aim for has perfected the mixture of the two, propose ‘global digital library’ as a con- ease of use. She says, “The audience while maintaining the generally high struct to encompass digital libraries for digital libraries has changed radi- quality of the presentations. This that are connected to, and accessible cally since the early days of informa- makes the NASIG proceedings one of through, a global information infra- tion retrieval, from expert search the few library science annuals actual- structure. A global digital library intermediaries to ‘every citizen’ who ly worth reading every year.—Bob would not be a single entity, nor would has access to the network. The next Persing ([email protected]), it be controlled by any single organiza- generation of digital libraries must University of Pennsylvania Libraries, tion” (52). The words are straightfor- serve a large and diverse community Philadelphia ward and clear; the ideas are and provide a large and diverse collec- compelling and reflect with ease the tion of information resources” (141). From Gutenberg to the Global diversity—some might say chaos— Research and development can do Information Infrastructure: that now characterizes our profession- more than address today’s problems. Access to Information in the al environment. Chapter 7 outlines “a research agenda Networked World. By Christine In the first four chapters, for making the next generation of dig- L. Borgman. Digital Libraries and Borgman defines and explains the ital libraries better suited to people’s Electronic Publishing. Cambridge, basic elements of the GII (Global information-related behavior in work, Mass.: The MIT Pr., 2000. 324p. Information Infrastructure), and education, and leisure contexts” $42 (ISBN 0-262-02473-X). explores relationships among them: (166–67). Four trends provide the An outpouring of titles about the “D” chapter 1, “The Premise and the skeleton around which research word—digital libraries, digital publish- Promise of a Global Information should develop: (1) availability of full- ing, digital technologies, digitizing for Infrastructure”; chapter 2, “Is It Dig- text databases, not just bibliographic preservation, digital materials, etc.— ital or Is It a Library? Digital Libraries data, and (2) linked, not stand-alone, gluts the professional literature. and Information Infrastructure”; systems; (3) systems that require users Readers have a right to ask why we chapter 3, “Access to Information”; to navigate, not just perform simple need this book too, which is about dig- and chapter 4, “Books, Bytes, and searches; and (4) concentrating on ital libraries even if it doesn’t use the Behavior.” As their titles suggest, group processes and social context, not “D” word in its title. The answer is chapter 1 defines what is meant by GII merely the individual user’s search twofold: first, it is eminently readable; and explains how it might be expected behavior. Borgman sees a “fundamen- and, second, it takes a step back from to function; chapter 2 describes digital tal challenge of balancing the need for 45(1) LRTS Book Reviews 61

tailoring digital libraries to a commu- ing any easy answers. Instead, she asks and Martha Yee. The rest are closely nity of users with the need to construct that, while we may not know exactly associated with cataloging or Lubetzky: a vast, interoperable, global digital where we are going, we pay attention John Byrum, Allyson Carlyle, Maurice library” (168). to what we already have learned about Freedman, Michael Gorman, Michael The final three chapters suggest where we are, and use it in deciding Malinconico, and Margaret Maxwell. ways of meeting the challenges of the how we might set about achieving Together, they cover a broad spectrum, future, including the possible imple- new, productive results. That shouldn’t though not all, of the theoretical terri- mentation of the GII. Chapter 7, be too much to ask.—Sheila S. Intner tory in the world of cataloging. “Whither, or Wither, Libraries?” is an ([email protected]), Simmons College Part 1 consists of three chapters. examination of the issues for libraries, Graduate School of Library and The papers focus on Lubetzky and the and chapter 8, “Acting Locally, Information Science, Boston origins of Anglo-American cataloging: Thinking Globally,” is a similar discus- “The Vicissitudes of Ideology and sion of successful, contributing digital The Future of Cataloging: Insights Technology in Anglo-American library initiatives. In chapter 9, from the Lubetzky Symposium, Cataloging since Panizzi” by Lubetzky “Toward a Global Digital Library: April 18, 1998, University of and Svenonius; “Seymour Lubetzky, Progress and Prospects,” Borgman California, Los Angeles. Ed. Man of Principle” by Gorman; and speculates on what may lie ahead, con- Tschera Harkness Connell and “Musings on Cataloging and cluding, “Among the greatest chal- Robert L. Maxwell. Chicago: Information Science” by Cloonan. lenges faced in realizing a GII is to ALA, 2000. 184p. $65 (ISBN 0- Lubetzky and Svenonius highlight the scale the present-day Internet to a 8389-0778-4). LC99-87247. two elements guiding our cataloging much larger, more complex, and more This slim volume, like Gaul, is divided traditions—the objectives of the cata- diverse information infrastructure that into three parts: “The History of log, which Lubetzky reinterpreted, will support many more users” (267). Cataloging and the Contributions of based on the turn-of-the-century That is a sobering thought for readers Seymour Lubetzky,” “Current Re- statement by Charles A. Cutter, and who believed the Internet was too search in Cataloging,” and “The technology. Gorman, as always, pro- much to handle. Future of Cataloging.” A small number vides readers with a literary delight as More than forty pages of refer- of pages is devoted to expected and well as a piece worth reading for pro- ences document the author’s sources appropriate sentiments expressed for fessional reasons. He bridges the gap and furnish a bibliography of relevant the honoree, who was about to cele- between Lubetzky’s adherence to the material for the reader interested in brate his one-hundredth birthday at principle of authorship and its misin- deeper exploration of these themes. the time the symposium was held. terpretation as being the same as main The listing is international and draws These include a photograph of entry. The book is worth buying and on the literatures of several disci- Lubetsky receiving an honorary degree reading for this chapter alone. plines, including computers, human from the University of California at Cloonan discusses the ideas of behavior, social science, telecommuni- Los Angeles (UCLA), excerpts from Vannevar Bush and the similarities cations, and more, in addition to the letters pertaining to his career, com- and differences from Lubetzky’s work familiar areas of library and informa- mentaries on his works, a partial bibli- in a chapter that opens the way to tion science. While welcome, it lacks ography of his publications, and a deeper exploration of the world of dig- the usual formatting, such as italicized clever poem by one of his colleagues. ital information, expanded in subse- titles or indented lines, and is hard to Fifteen authors, including the hon- quent papers in the book. read and use. The book closes with a oree, contributed fourteen papers (the Part II contains five chapters: detailed index. first is a collaboration of Lubetzky and “Modeling Relevance in Art History” From Gutenberg to the Global Elaine Svenonius). One could easily by Layne; “Creating Efficient and Information Structure is highly rec- anticipate the authors represented Systematic Catalogs” by Carlyle; ommended to all its potential read- here, since most are associated with “Main and Added Entries” by ers—librarians and information UCLA: Marcia Bates, Michael Carpenter; “Lubetzky’s Work Prin- specialists, technologists, network spe- Carpenter (a UCLA Ph.D., now teach- ciple” by Yee; and “Applying the cialists, policymakers, scholars, pub- ing at Louisiana State University), Concept of the Work to New lishers, and other information Michèle Cloonan, Sara Shatford Environments” by Leazer. Although industrialists. Borgman offers a well- Layne, Gregory Leazer, Svenonius, all five illuminate important aspects of focused assessment of likely areas of Barbara Tillett (also a UCLA Ph.D., research in areas in which Lubetzky contention and problems in the evolu- now director for the Integrated Library pioneered study, Yee’s paper is by far tion of a GII, and wisely eschews giv- System at the Library of Congress), the most intriguing. After giving a 62 Book Reviews LRTS 45(1)

thorough and fascinating report of her some material for subsequent follow- by human indexers. The computer sci- research into certain characteristics of up, though references to AACR ence approach, on the other hand, has main entry and the works they repre- abound. Unfortunately, the print is concentrated on “content-based” sent, she summarizes: “One of small and the margins narrow, indicat- analysis that relies on automated pro- Lubetzky’s gifts to library users who ing that the book was not intended to cessing of the images themselves. It is seek particular works was his explica- withstand heavy use or be rebound good to see in this compilation that tion of the work principle, which has after many readings and rereadings. most of the authors now view the two the potential to allow us to design More is the pity, because this book is approaches to image analysis as com- OPACs [online public access catalogs] worth understanding by the masses— plementary rather than as an either-or that meet the cataloging objectives myriad practitioners working with choice. better than any catalogs we have ever online catalogs, which remain the The three papers in the first sec- seen before. The generations of library library’s basic tool for patron service. tion, “Foundations of Access to Visual leaders that followed Lubetzky Because the core of the text is Information,” are devoted to the prob- dropped the ball, however, and focused on research, this book promis- lems of access to images from three allowed the development of OPACs es to have an active, useful life for the different perspectives. Hsin-liang that impede the user who seeks partic- foreseeable future. It will always have Chen and Edie M. Rasmussen, in ular works much more than the card historical value and be of interest for a “Intellectual Access to Images,” catalog ever did” (102). That should while for its prognostications; but the describe briefly the tools currently give pause to the people who design likelihood that practical payoffs from available for providing intellectual online catalogs and, if they are made the research reported here will take access to images and the contexts in aware of this and other studies, years to realize means that the influ- which these tools are being used. P. prompt them to begin thinking of cat- ence of the Lubetsky Symposium Bryan Heidorn, in “Image Retrieval as alogs as intellectual rather than should remain fresh and vital for quite Linguistic and Nonlinguistic Visual mechanical devices. some time.—Sheila S. Intner (shemat@ Model Matching,” explores the com- The third and final part consists of aol.com), Simmons College Graduate plexities of visual perception and six chapters: “The Ideology and School of Library and Information memories and describes ways in which Technology of Cataloging at the End Science, Boston systems might be designed to assist of the Millennium” by Freedman; users in retrieving images that match “Cataloging Virtual Libraries” by Progress in Visual Information the mental models of the images they Malinconico; “World Wide Web Access and Retrieval. Ed. Beth are seeking. Heidorn focuses on users Opportunities in Subject Cataloging Sandore. Library Trends 48, no.2 who have mental models of the images and Access” by Bates; “Cataloging at (fall 1999): 283−524. Champaign: that they are seeking, but does not Crossroads: Preservation and Accom- University of Illinois Graduate address the needs of users who may be modation” by Byrum; “Guidelines for School of Library and Infor- seeking images because they do not a Future Anglo-American Cataloging mation Science, 1999. Single know what something looks like. In Code” by Margaret Maxwell; and issue, $18.50 (ISSN 0024-2594). “Computer Vision Tools for Finding “Current Activities in Cataloging Code Sandore’s stated goal in this compila- Images and Video Sequences,” D. A. Revision” by Tillett. Again, each of tion of papers is to present a diverse Forsyth concentrates on automatic these papers is worth time and audience “with a current perspective tools for providing access to images, thought. They move from the theoret- on the development of visual informa- describing their current limitations ical plane to the practical, describing tion retrieval and access tools” (283). and suggesting ways in which these potential changes to the Anglo- She succeeds admirably in accom- limitations might be overcome. American Cataloguing Rules (AACR) plishing this goal, providing not a sin- The second section, “Imple- in the context of goals to be achieved gle perspective but rather several mentation and Evaluation,” is richer in and the new environment in which a perspectives on a variety of issues. The content than its title might imply. The cataloging code must function. authors of the papers come from dif- papers in this section are not simply of The book is well edited and sim- ferent backgrounds, encompassing the show-and-tell variety, but rather ply, but neatly, formatted. Editors both library science and computer sci- incorporate thoughtful analysis of the Connell and Maxwell did not con- ence. In the past, the library science often-complex issues surrounding tribute original papers to the volume, approach to providing access to access to images. Practical guidance but their work is evident in the smooth images has concentrated on “concept- may be found here, but always in the flow of language and lack of errors. based” analysis of images that relies on context of theoretical issues and users’ Notes at the ends of chapters offer textual terminology usually assigned needs. Teresa Grose Beamsley, in 45(1) LRTS Book Reviews 63

“Securing Digital Image Assets in that all three of these approaches the good who praise their efforts to Museums and Libraries: A Risk make use of textual information as well prove that “information is inevitably Management Approach,” discusses as automated analysis of the image embedded in social relations” (Robert the details of digitizing images but at itself. Yong Rui et al., in “Information D. Putnam, 3) and that “information the same time explores the problems Retrieval beyond the Text Document,” technology does not work unless sup- of preserving image integrity and describe an experimental technique ported by viable communities and image context. Beamsley also address- tested in the context of images of institutions” (Bruce Kogut, 3). Even a es some of the issues surrounding museum objects. They focus on what cursory reading of this text demon- copyright and control of images in the the computer can do best—provide an strates that Brown and Duguid are digital world. In “Getting the Picture: analysis of color, texture, and shape— serious fellows appropriately skeptical Observations from the Library of and suggest that this analysis is most of the prevalent cyberhype but, at the Congress on Providing Access to successful when used in conjunction same time, convinced of the transcen- Pictorial Images,” Caroline R. Arms with manually supplied, text-based dent importance of the changes that describes the processing of images in access. Both Neil C. Rowe in “Precise technology is wreaking and will wreak the vast collections of the Prints and and Efficient Retrieval of Captioned in the future. So far, so very good— Photographs Division of the Library of Images” and Rohini K. Srihari and and the fact that the authors write in Congress. She details all aspects of Zhongfei Zhang in “Exploiting acceptable English prose is another processing, from specifications for dig- Multimodal Context in Image unexpected plus. itizing images to the design of the Retrieval” describe experiments in I am a simple and prosaic soul and retrieval interface. Arms’s paper is par- which automated text processing is always approach books of this type ticularly valuable because she combined with image processing in with two simple and prosaic tests. describes methods for processing a order to retrieve images from Web How do the authors define informa- large collection of images, including pages. All three papers in this section tion? What do you find when you look the creation of collection-level and include detailed descriptions of the up library or libraries in the index? group-level records, and for integrat- methodology used in the automated (The latter on the grounds that how ing access to images with access to analysis. authors treat something that I know other materials, as occurs, for exam- In summary, the papers in this well is a fair guide to how they treat ple, in the Library of Congress’ online compilation provide a theoretical what is less well known.) It is a sad fact catalog. Christie Stephenson, in exploration of issues, a description of that many books on information, the “Recent Developments in Cultural the current state of access to images, Information Age, and so on fail to pro- Heritage Image Databases: Directions some practical guidance, and a look at vide any definition of information, or for User-Centered Design,” describes possibilities for the future. This com- give en passant a cursory and practi- the experience of the Museum pilation should be useful to both cally useless definition. My spirits lift- Educational Site Licensing Project librarians and computer scientists and ed when I found a discussion of the (MESL) and in the process explores to anyone involved in providing access differences between knowledge and the issues involved in providing access to image collections.—Sara Shatford information (117–19), but not for long, to images from museum collections. Layne ([email protected]), because apart from the smart-sound- In “Evaluation of Image Retrieval Science and Engineering Library, ing “knowledge usually entails a know- Systems: Role of User Feedback,” University of California, Los Angeles er” (119), the discussion ends with Samantha K. Hastings describes a thrown up hands. I write “smart- study performed using a relatively The Social Life of Information. By sounding” because the implication is small and homogeneous set of users John Seely Brown and Paul that information can exist without and images. Based on this study, Duguid. Boston: Harvard Business human consciousness—an interesting Hastings categorizes user queries and School Pr., 2000. 320p. $25.95 philosophical question but one with- suggests the types of tools that are best (ISBN 0-87584-762-5). LC99- out practical utility. Surely, even data suited to address these queries, as well 49068. without a human mind is unbearably as the best ways of evaluating the suc- The authors are, respectively, the evanescent. There is a definition of cess of these tools. director of the Xerox Palo Alto sorts for information: “something that The final section, “Experimental Research Center and a historian and can be recorded, transcribed, digi- Approaches,” consists of three experi- social theorist research specialist at the tized, and shipped in packets” (2). mental automated approaches to pro- University of California-Berkeley. Apart from the leadenness of the viding access to images. It is an They come garlanded with encomia words, the fact that they can apply to interesting and welcome development from a variety of the digital great and everything from a Mozart symphony 64 Book Reviews LRTS 45(1)

to Citizen Kane to a Harry Potter book nological change. What they call the the ways in which they were to a line of Census data to a street map “limits to [sic] information” (vii) center “infofriendly” (117), while ignoring of Duluth makes the definition of no on the notion that technology is a tool wider human and societal issues. Any utility at all. The result of the second to be used or misused in societies and librarian of a certain age will enjoy the test, looking up “Libraries, digitized, human lives but without the extreme authors’ analysis of the myths of 179−181”—the only such entry in an consequences foreseen by the wilder- paperlessness. “Infoenthusiasts have exiguous index—led to even more eyed digital pundits. The authors then thrust nothing under the hammer with dispiriting words. The “conventional move on to discuss the idea of agents, quite so much enthusiasm as the paper library, with its massive weight of in the nauseating nomenclature of the document” (174). They go on to show paper gathering dust and resisting effi- day—knobots, chatterbots, shopbots, how and why the paperless office, the cient searches, is another paper-based etc.—and their limitations, which are electronic newspaper, and the digital institution that sets fingers itching at many in number and kind. They point library have all turned out to be illuso- the keyboard. The sense that the infor- to the way in which such agents are ry. In the chapter titled “Re-educa- mation is ‘there’ somewhere, but can’t given human names (Jeeves, Sherlock, tion” they address the future of be found can drive anyone to digitize” Bob, and the like) as an indication that universities and puncture many of the (179). Seldom can there have been their advocates wish to narrow the dis- myths contained in the increasingly two sentences more choc-a-bloc with tinction between humans and soft- voluminous and silly writings on the clichés, straw men, and plain drivel. It ware, and would seek to replace topic. Of interest to us, the authors, ill behooves two men bent on counter- humans with agents whenever possi- unlike most academics, seem to acting digital hype to claim that infor- ble. One would have thought this self- understand that the future of the uni- mation can’t be found in conventional evidently negative and potentially evil, versity library is far more likely to libraries. If the Web and the Net had but the authors believe that there are resemble its past than it will the opium one one-hundredth of the organiza- areas to be negotiated and demarcated dreams of futurists, information “sci- tion and retrieval architecture of a to protect us from antisocial agents. In entists,” and renegade librarians. conventional library, the world would their words, “[w]e want ATMs that will In sum, this book offers a good be a happier, saner place. supply us with money, but not ones analysis of the blinkered nature of most This book is better than its lack of that will spend it for us” (62). prognostications on technology and a definition of its subject and stereo- In subsequent chapters, Brown “information,” and of the way those typing of libraries would indicate. The and Duguid explore the isolating prognostications ignore the social and authors begin with an admirably clear- effects of technology on the individual human context of communities, organ- eyed assessment of the overheated and the consequences for the way in izations, and institutions. The authors propaganda on behalf of technology to which enterprises are organized. They do not claim to offer solutions to the which we are all subject. Their central cast a cold eye on the discredited and intellectual mess we are in, but they do metaphor is one of tunnel vision—a antihumanistic management fads of offer a way of looking at that mess that concept that envisions the technologi- the 1990s—downsizing, reengineer- may well lead to solutions.—Michael cal drive into the future as being nar- ing, etc., and the fatuous TQM, of Gorman ([email protected]), row and exclusionary in that it ignores which far less is heard since the Henry Madden Library, California the human and social context of tech- Japanese economy crash-landed—and State University, Fresno 45(1) LRTS 65

Instructions for Authors

Manuscript Submission acronyms. All acronyms must be 9. Be prepared to supply camera- accompanied by their full ready copy for all illustrations. Manuscripts of articles should be sent spelled-out form. For spelling Accompany the manuscript to the editor, John Budd, University of and usage, consult the Random with a photocopy of each and a Missouri, 221M Townsend Hall, House Webster’s College Diction- brief, meaningful caption noted Columbia, MO 65211; (573) 882- ary (New York: Random House, on the verso. 3258; fax: (573) 884-4944; e-mail: 1991). Verify the spelling and [email protected]. accuracy of all names in an In general, the editorial staff fol- appropriate source. Consult The Editorial Policy lows the Guidelines for Authors, Chicago Manual of Style 14th ed. Editors, and Publishers of Literature (Chicago: Univ. of Chicago Pr., Library Resources & Technical Services in the Library and Information Field 1993) for capitalization, abbrevi- (LRTS) is the official journal of the adopted by the American Library ations, usage of numbers, etc. Association for Library Collections & Association Council in 1983 and 4. Give the article a brief title; if Technical Services (ALCTS), a division available from the ALA Executive the title does not fully describe of the American Library Association. Offices. Information about copyright the content of the article, add a The following statement of editorial policies also is available from ALA brief subtitle. On the first page policy was adopted by the ALCTS headquarters. of the manuscript give the arti- Board of Directors, April 1998. cle title, the name(s) of the author(s), and the position title, Purpose Manuscript Preparation institutional affiliation, and address of each author. The purpose of LRTS is to support Please follow these procedures for 5. On the second page of the man- the theoretical, intellectual, practical, preparing manuscripts for Library uscript give the title followed by and scholarly aspects of the profes- Resources & Technical Services (LRTS): a brief, informative abstract. Do sion of collection management and not identify the author(s) here development, acquisitions, cataloging 1. Submit original, unpublished or elsewhere in the manuscript. and classification, preservation and manuscripts only. Do not sub- Number all pages throughout reformatting, and serials, by publish- mit manuscripts that are being the manuscript. ing articles (subject to double-blind considered for publication in 6. Submit all references on sepa- peer review) and book reviews, and other venues. Authors are rate pages at the end of the editorials and correspondence in responsible for the accuracy of text, preceding any tables or response to the same. statements included. Papers illustrations. presented at a conference 7. LRTS follows The Chicago Audience should be identified with the Manual of Style author-date conference name and date in system of references (see chap- The audience for LRTS includes prac- the cover letter. ter 16). Verify each citation by titioners, students, researchers, and 2. Manuscripts may be submitted sight, very carefully. other scholars with an interest in col- on paper (three copies, machine- 8. Follow the examples and sug- lection development and technical printed and double-spaced) or in gestions in chapter 12 of The services and related activities in all an electronic form (Word file Chicago Manual of Style in types of libraries. attached to e-mail). Disk copy designing tables. Submit each will be requested from authors table on a separate page at the Frequency for accepted articles. end of the manuscript. Indicate 3. Write the article in a grammati- the preferred placement in the LRTS is published quarterly, with the cally correct, simple, readable text with an instruction in square volume calendar corresponding to the style. Whenever possible avoid brackets. Provide each table calendar year. Numbers appear in jargon, anthropomorphism, and with a brief, meaningful caption. January, April, July, and October. 66 Instructions for Authors LRTS 45(1)

Scope technical services. The scope of the 3. Notes that report unique or articles published in LRTS is also guid- evolving technical processes. The editor of LRTS, with the assistance ed by the Mission and Priorities 4. Notes that report unique or of an editorial board, strives to achieve Statement adopted by the ALCTS evolving research methods. a balance among the articles published Board of Directors in 1990. 5. Substantive book reviews of in the journal so that the interests of new publications. each of the sections of ALCTS Content 6. A brief, factual, annual state- (Acquisitions, Cataloging and Classifi- ment of the association’s accom- cation, Collection Management and The content of LRTS is to include: plishments. Development, Preservation and Refor- matting, Serials) is represented in the 1. Articles that further the LRTS is not an appropriate forum journal. Articles on technology, man- advancement of knowledge by for brief reports on new products, agement, and education, e.g., are reporting the results of research new services, or other current news appropriate to the journal when the or other scholarly activity. items. application of these is to issues of inter- 2. Periodic literature review est to practitioners and researchers essays that discuss issues and working in collection development and trends. 45(1) LRTS 67

Instructions for Book Reviewers

Length the author-date system of references, ● do not use your word proces- as described in The Chicago Manual sor’s automatic endnoting capa- Manuscripts should be 800–1,000 of Style, chapter 16. Please verify all bility; words (3–4 double-spaced pages). citations carefully. ● treat the endnotes as the final Reviewer’s name, e-mail address, paragraph of your review; Content and affiliation immediately following ● include at least 1" margins on the text: if unaffiliated, give city of res- all sides. Reviews should evaluate books as a idence. contribution to the literature of Do not use end notes for reasons Submitting the Review library and information science, char- other than the citation of other works. acterizing the content of the book and Use first name or initial of each If possible, please send your review as critically appraising it. Try to position person cited for the first time. an e-mail attachment, preferably a the book within the intellectual histo- Avoid library jargon because it Microsoft Word Document. If e-mail ry of the field by comparing it to other may become dated (e.g., online cata- is not possible, send the review on a books or ideas that are related to it. log, not OPAC). 3.5" disk. Library Resources & Technical In quotations, identify acronyms Services is not a library book selection and initialisms (e.g., ALA, ALCTS, And . . . A Few More Suggestions, tool and does not include purchasing ASIS, etc.) with the full name at their Especially for Novice Reviewers recommendations. first use: “Library of Congress (LC) policy is.” Use the acronym for subse- Allow enough time to read the book, Style quent references to the same entity. think about it, draft your review, and For good advice and helpful rewrite. Don’t base your review on a LRTS follows The Chicago Manual of reminders on writing and style, check careful reading of the book’s introduc- Style, 14th ed. (1993). Please include the following compendium, which tion and conclusion or—heaven for- the following information in the includes a link to the 1918 edition of bid—the book jacket and other review: Strunk’s The Elements of Style: reviews. It is often helpful to ask a col- Header: complete bibliographic http://andromeda.rutgers.edu/~jlynch league to read your review and com- citation of the book, in the following /Writing. ment. If you would like to send a draft format: for comment before preparing the Format final version, feel free to do so, well in Advances in Library Auto- advance of the deadline. mation and Networking, Vol. Reviews submitted in machine-read- A review is not an abstract of the 5 (1994). Ed. Joe A. Hewitt able form are greatly appreciated. To book. To characterize the book’s con- and Charles W. Bailey, Jr. facilitate editing, please observe these tent, a synthesis or a description of its Greenwich, Conn.: JAI Pr., additional formatting instructions: highlights is preferable to a chapter- 1994. 282p. $73.25 (ISBN 1- by-chapter summary. This is especial- 55938-510-3). ● double-space all text, including ly important, and also a difficult task if headers; you are reviewing a compilation. Quotations: give page numbers in ● do not indent paragraphs, but Careful attention to this part of the parentheses (36) (vii) immediately fol- use two carriage returns review usually will help you focus on lowing the quote. between paragraphs; evaluating the book. Quotation marks around chapter ● do not use carriage returns at A review is an opinion piece, not or section heads: be sure these reflect line ends (use your word proces- a book report. State opinions clear- the exact language used in the book. sor’s word-wrap capability); ly—be fair and specific with both Author(s), title, place, and date of ● keep the right margin ragged praise and criticism. Understand the publication of all works cited: a com- (do not use full justification) author’s purpose in writing the book plete bibliographic citation should be and do not hyphenate words at and evaluate the success in meeting included at the end of the text. Use the line breaks; stated objectives. Are the objectives 68 Instructions for Book Reviewers LRTS 45(1)

worthwhile? How does the book com- And One Suggestion review style or about submitting your pare to others of similar content and for Everyone . . . review, please contact me: Margaret purpose? What is unique or especial- Rohdy, Head of Research, Training & ly valuable about the book? Is the Clear your mind, especially at the Quality Management, University of book controversial? To what audience rewrite stage, by taking another look at Pennsylvania Libraries, 3420 Walnut will it be most interesting and useful? Strunk and White. Your editor does! St., Philadelphia, PA 19104-6206; Are there any significant errors or Thank you for reviewing a book for phone: (215) 898-0314; fax: (215) 573- omissions? LRTS. If you have questions about 9610; e-mail: [email protected].